From Drones to Geospatial Data

Dr. Jean A. Doumit

Reviewed by Prof. Dr. Anatoly V. Pogorelov

Lebanese University
Kuban State University
 Acknowledgements

This book is dedicated to all Geomatics Engineers and Earth Scientists,

the following organizations are thanked for permission to use photographs and
data:
The Lebanese army, the Lebanese University, Kuban State University,
SOUTH company for surveying Instruments and Aerogeomatica.
Three must be singled out for special mention: Professor Pogorelov Ana-
toly Valerievich as the best scientific advisor, the department of Geoinformatics
at Kuban State University, and the department of Geography at the Lebanese
University.
I would also like to express my gratitude to all who helped me take this
book from a vision to an actualization, especially my parents for their constant
support.
Finally, my Very Special Thanks go to my dear wife, for her forbearance
in accepting the disruption of our social life and for her encouragement, feeding
and watering me, during the preparation of this book.

2
 Preface

Unmanned aerial vehicles were once the stuff of rumor and legend, iden-
tified as new and Mysterious robots in the sky.
Apart from military applications there are many jobs to be performed
monitoring and rescue, the unmanned aerial vehicles are used for photogramme-
try tasks.
This new technology been the case across various disciplines such as
mapping, remote sensing, civil engineering, geology, geomorphology, military
engineering, land planning, and communications. It is encouraging that more
literature is now available in this discipline. This book was written with the in-
tention to add to and benefit this field of study.
I am pleased to present you with this 1st edition of this book. It introduces
the classes and types of platforms available, and examines their performance,
structure, propulsion and systems to be suitable for photogrammetry. Sensor and
payloads, methods of launch and recovery, image processing and geospatial
analysis.
In its nine chapters, this book carries out several tasks: Drones field of
uses, classification of drones, technical structures of drones, drones flight plan-
ning and software, drone photogrammetry methods, data processing and geospa-
tial analysis. The purpose of this book is to give an idea about drone for re-
searchers and students to apply this new technology in researches and projects,
and introduce them to the Unmanned Aerial Vehicles new era leading to a treas-
ure of information. My message to the readers is: “Use new technology”.

3
Table of Contents
Chapter 1 Introduction to Photogrammetry ............................................................................... 5
What are a UAV and UAS? ................................................................................................... 7
Description of a drone ............................................................................................................ 8
Drones field of Uses ............................................................................................................... 9
Chapter 2 Advantage and limitation of Drones ........................................................................ 11
Advantages of drones ........................................................................................................... 11
Limitations in the use of drones ........................................................................................... 12
Chapter 3 Classification of drones ........................................................................................... 13
Fixed wings .......................................................................................................................... 13
Rotary systems ..................................................................................................................... 14
Chapter 4 Drones structures ..................................................................................................... 19
Unmanned Aerial Vehicles (UAV) ...................................................................................... 19
Unmanned Aircraft System (UAS) ...................................................................................... 21
Chapter 5 Drone Software’s ..................................................................................................... 26
Drones flight planning software ........................................................................................... 26
Drones mapping software..................................................................................................... 28
Processing workflow ............................................................................................................ 30
Chapter 6 Drones photogrammetry methods ........................................................................... 34
Type of photogrammetry ...................................................................................................... 34
Close range photogrammetry (CRP) .................................................................................... 34
Distortions and Camera calibration ...................................................................................... 38
Bundle adjustment ................................................................................................................ 41
Structure from Motion (SFM) .............................................................................................. 43
Aerotriangulation ................................................................................................................. 46
Chapter 7 Flight Planning Principles and image processing .................................................... 48
Ground sampling distance (GSD) ........................................................................................ 48
Photocontrol points .............................................................................................................. 49
The Global Positioning System (GPS) ................................................................................. 51
Flight planning ..................................................................................................................... 53
Image processing .................................................................................................................. 55
Drone mapping accuracy ...................................................................................................... 57
Ground Sampling Distance (GSD) accuracy ....................................................................... 58
DSM Accuracy Assessment ................................................................................................. 58
Chapter 8 Drones and Geospatial Analysis .............................................................................. 59
Multi-scale Digital Surface Models ..................................................................................... 61
Multi-scale landforms classification .................................................................................... 68
Terrain analysis for parcels evaluation................................................................................. 77
Multi-scale Terrain Roughness ............................................................................................ 83
Digital Surface Models to planar areas. ............................................................................... 89
Chapter 9 Drones regulations and buyers guide....................................................................... 98
Drones Regulations .............................................................................................................. 98
Flight safety .......................................................................................................................... 98
Drones buyers guide ........................................................................................................... 100
Conclusion .............................................................................................................................. 101
References .............................................................................................................................. 102

4
 Chapter 1 Introduction to Photogrammetry

This book provides a general overview of drone photogrammetry, its theo-

ry and general working principles with an emphasis on concepts not on mathe-
matical formulas and equations.
Photogrammetry is an engineering discipline and as such heavily
influenced by developments in computer science and electronics.
Photogrammetry, from the Greek photo (light writing) gram (graphics)
metry (measure), it is “the art, science, and technology of obtaining reliable in-
formation about physical objects and the environment, through processes of re-
cording, measuring, and interpreting images and patterns of electromagnetic ra-
diant energy and other phenomena”.
Photogrammetry produces maps directly from photographic images by
identifying, symbolizing, and compiling elevation, cultural, and natural features
that are visible on the imagery.
Each specialist use photogrammetry for his purposes:
 A soils specialist may be looking for erodible soils on a proposed
highway route.
 A forester may be estimating the volume on a timber tract.
 A hydrologist may be comparing the degree of suspended matter in
several adjacent lakes.
 A wetland specialist may be monitoring the decline of wetland areas
within a state.
 A glaciologist may be charting the movement of a glacier, etc.
Several types of photogrammetry exist: aerial, terrestrial, and close range.
Each serves the needs of a distinct category of users. Throughout the mapping
community, terrestrial and close-range photogrammetry has limited use. Aerial
photogrammetry uses near-vertical photographic images that are exposed to a
moving platform at a distant point in the sky (Falkner, Morgan, 2002).
Photogrammetry and remote sensing are two related fields, the principle
difference between photogrammetry and remote sensing is in the application;
while photogrammetrists produce maps and precise three-dimensional objects,
remote sensing specialists analyze and interpret images for deriving information
about the Earth’s surface. Both disciplines are also related to Geographic Infor-
mation Systems (GIS) in that they provide GIS with essential information.
Photogrammetry is the science of obtaining information about land sur-
faces without physical contact, for the purposes of measuring and interpreting
this information.
Figure 1 shows a logic systems simplify understanding the photogramme-
try procedure.

5
 Fig.1.1: The logic system for understanding photogrammetry.

Figure 1.1 shows the workflow of photogrammetry beginning from data

acquisition from sensors passing by the processing step or procedures and end-
ing by the data production.
The development of photogrammetry clearly depends on the general de-
velopment of science and technology. It is interesting to note that the four major
phases of photogrammetry are directly related to the technological inventions of
photography, airplanes, computers, and electronics.
Photogrammetry had its beginning with the invention of photography by
Daguerre and Niepce in 1839. The first generation, with remarkable achieve-
ments in terrestrial and aerial balloons.
The developments in photogrammetry, from around 1850, have followed
four development cycles.
Each of these periods extended about fifty years. These cycles include:
(a) Plane table photogrammetry, from about 1850 to 1900,
(b) Analog photogrammetry, from about 1900 to 1960,
(c) Analytical photogrammetry, from about 1960 to 2010,
(d) Digital photogrammetry, which just began to be a presence in the pho-
togrammetric industry.

6
 In 1855, Nadar (Gaspard Felix Tournachon) used a balloon at 80-meters
to obtain the first aerial Photograph, in 1859 Emperor Napoleon ordered Nadir
to obtain reconnaissance photography in preparation of the Battle of Solferino
(Gruner, 1977).
The English meteorologist E.D. Archibald was among the first to take
successful photographs from kites in 1882.
In France, M. Arthur Batut took aerial photographs using a kite, over
Labruguiere, France, in May 1888.
In 1893, Dr. Albrecht Meydenbauer (1834-1921) was the first person to
use the term "photogrammetry" (Meyer, 1987).
In 1903, Julius Neubranner, photography enthusiast, designed and patent-
ed a breast-mounted aerial camera for carrier pigeons.
1903: Airplane invented by Wright brothers.
1909: The Wright brothers take the first photograph from a plane over
Centocelli, Italy.
Captain Cesare Tardivo (1870-1953) is thought to be the first to use aerial
photography from a plane for mapping purposes. He created a 1:4,000 mosaic of
Bengasi in Italy that was described in his paper to the 1913 International Society
of Photogrammetry meeting in Vienna.
First drones investigations started at Carnegie Mellon University (Ei-
senbeiss, 2011).
(Nagai et al., 2004) used LiDAR system mounted on a drone, in 2008 they
showed the latest results for multi-sensor integration on drones.
After 2010 the evolution and the invention of new drones systems were
very fast and put a multidisciplinary powerful scientific engine among civilian
hands.

What are a UAV and UAS?

Unmanned Aerial Vehicles “UAVs are to be understood as uninhabited
and reusable motorized aerial vehicles” (Von Blyenburg, 1999). These vehicles
are remotely controlled, semi-autonomous, autonomous, or have a combination
of these capabilities.
The UAV is an acronym for Unmanned Aerial Vehicle, which is an air-
craft with no pilot on board. UAVs can be remote controlled aircraft or can fly
autonomously based on pre-programmed flight plans or more complex dynamic
automation systems.
The acronym UAV has been expanded in some cases to UAVs (Un-
manned Aircraft Vehicle System). The FAA has adopted the acronym UAS
(Unmanned Aircraft System) to reflect the fact that these complex systems in-
clude ground stations and other elements besides the actual air vehicles).
Officially, the term 'Unmanned Aerial Vehicle' was changed to “Un-
manned Aircraft System”. The term UAS, however, is not widely used as the
term UAV has become part of the modern lexicon.

7
 The use of professional civilian drones is increasing rapidly around the
world and is expected to explode in the coming years.
In the following chapters of our book, we will use the term "drone" in-
stead of UAV and UAS.

Description of a drone
A simple drone system has two basic parts, airborne part, aircraft frame
carrying the payload of the system and containing (frame, camera, battery, gim-
bal, etc.) and the ground-based part which constitute the operation base interface
of all the system (base station and a radio transmitter) in chapter 6 we will dis-
cuss it in details.
Some drones have the possibilities of automated take-off and landing
without any manual operator, few drones can be functioned without a link to a
Remote Control (RC) transmitter or base station. This constraint is implemented
mainly as a safety precaution to ensure that manual control of the UAV can be
resumed if needed at any given point during a flight.
Drones are categorized into fixed wing, and helicopters. These drones are
powered by brushless electric motors or by combustion engines. Electric motors
cause significantly fewer vibrations in the airframe than internal combustion en-
gines and are preferred for photographic applications.
It is not very difficult to learn how to fly RC helicopters or fixed wing in a
short time by it needs practice and training. Some vehicles are very easy to con-
trol and don’t need any flight experience and training because it contains auto-
mated orientation sensors and processing power keeping the platform aloft.
A basic flight control system for these drones contains very specific sen-
sors linked to a processor that manages power distribution to the motors to stabi-
lize flight. Most flight control systems are composed of the magnetometer, a ba-
rometer, and Global Positioning Systems (GPS) receivers to support three-
dimensional drones navigation.
In this book, we will give an overview of existing classifications, struc-
tures, compositions of these drones and their field of applications, especially in
geography.
The unmanned aircraft system (UAS) is a fully designed system compris-
es:
a) A control station (CS) constituting the operation base interface of all
the system.
b) Aircraft frame carrying the payload of the system.
c) A communication system between the CS and the aircraft and contrary,
this system is achieved by radio transmission (Austin 2010).

8
 Drones field of Uses
Drones photogrammetry opens various new applications at a very low-
cost comparing to the classical photogrammetry, it allows to capture high-
resolution and to process data manually or automatically.
Drones are used in a numerous very huge number of fields:
Agriculture, drones provide an efficient way to assess plant health, after a
flight drone images are integrated into a special software for the generation of
index maps showing clearly, where plants are struggling or vegetation monitor-
ing (Sugiura et al., 2005). Lidar systems mounted on drones to measure the
height of crops, multi-spectral instruments can count plants, infrared imagery
can check the health of plants etc...
Archeology, drones used for the detection and mapping of archeological
sites and monuments (Bendea et al., 2007; Patias et al., 2007).
Architecture, simultaneous as built and scanning of the projects made by
drones can help in faster design and inspections, quality control, make meas-
urements and modifications.
Civil engineering, this field benefits from inspecting infrastructure,
bridges, cell phone towers, dams, powerlines, solar fields and road planning and
designs.
Crime analysis, drones can help in crime detections and investigations.
Customs, drones can surveil coastline, ports, bridges and other access
points for the import of illegal substances is the remit of the customs (Surveil-
lance for illegal imports)
Electricity Companies, inspections of power-lines, damage of structures
and deterioration of insulators.
Environmental impact assessment, the use of drones for glacial model-
ing to animal’s movement study, coastal erosion, etc.…
Fire Services, Fire departments can use drones to track and map wildfires
or for forest fire monitoring (Zhou et al., 2005).
Fisheries, the prevention of illegal fishing aided by patrolling drones
Forestry, drones can take a forest control by trees counting, trees health
and fire controls.
Gas and Oil Companies, use drones for pipes patrolling to look for dis-
ruption or leaks in accidents.
Geology, drones with specific electromagnetic sensors, can be used to
gather geological information to approximate the location and presence of min-
erals, oil, and natural gas.
Humanitarian aid, Drones are being increasingly used by NGOs and
governmental organizations to respond to and assess the impact of natural disas-
ters. For Rescue, drones with infrared sensors can be used to detect humans that
is helpful in search scenarios.

9
 Information Services and Mass Media, television companies, newspa-
per publishers could have the means of covering events, whether planned or ac-
cidental. Sports events could be covered in real-time.
Land surveying, Land surveyors use drones to survey inaccessible areas
like cliffs, rivers, woods etc… to create geo-referenced maps, 3D models, Digi-
tal Surface Models (DSM) and other data products.
Land registry, drones have made it possible to count buildings and popu-
lation growth estimation.
Meteorological Services, use drones to sample the atmosphere for fore-
casting by integrating special sensor.
Mining, Drone data can be used for taking measurements, automated
workflows create orthophotos as well as 3D models that can be used to quickly
calculate quantities for materials such as aggregate stockpiles or cut and all.
Photomontage, drones are used for posters making and cinemas filming.
Rivers Authorities have used drones successfully in monitoring water-
course flow and water levels. Rivers Authorities Water course and level moni-
toring (Masahiko, 2007).
Traffic monitoring (Haarbrink and Koers, 2006, Puri, 2004), road fol-
lowing (Egbert and Beard, 2007), vehicle detection (Kaaniche et al., 2005), car
accident and flare inspection of an industrial flue (Haarbrink and Koers, 2006).
Geo-informatics, Drones are increasingly being used in place of satellite
and photogrammetry images for the creation of up-to-date, high-resolution base
maps Geospatial data collection with high geometric and temporal resolution.
Below a small comparison between drones, photogrammetry and satellite
images.

Table 1.1: A brief comparison between satellite images, photogrammetry, and drones.

Space-borne Airborne Drones-borne

1. Extensive coverage. 1. Large coverage 1. Very high resolution
2. Low resolution (30 2. High resolution 2. (Fixed wing 2 cm/pixel,
cm/pixel). (7cm/pixel). rotary sub-millimeter).
3. Timing controlled by the 3. Timing controlled by the 3. Imagery acquired on de-
provider. provider. mand.
4. Cloud cover. 4. Susceptible to weather. 4. Susceptible to weather.
5. Expensive. 5. Expensive. 5. Unaffected by cloud cover.

The difference between satellite images, ordinary photogrammetry and

drones photogrammetry listed in table 1.1 could be summarized in: Satellite for
small scales, photogrammetry for big scales and drones photogrammetry very
big scales.

10
 Chapter 2 Advantage and limitation of Drones
Advantages of drones
Main drones advantages compared to manned aircraft systems it can be
used in high-risk situations without endangering a human life and inaccessible
areas. These regions are for example natural disaster sites, floodplains, earth-
quake and desert areas and scenes of accidents. Furthermore, in cloudy and driz-
zly weather conditions, the data acquisition with drones is possible, such weath-
er conditions do not allow to fly into manned aircraft.
Moreover, drones have the real-time capability and the ability for fast data
acquisition, while transmitting the image, video and orientation data in real time
to the ground control station.
Most of the (non-)commercially available UAV systems on the market fo-
cus on low-cost systems, and thus a major advantage of using UAVs is also the
cost factor, as UAVs are less expensive and have lower operating costs than
manned aircraft have. But, sometimes depending on the application the cost can
be similar to manned systems. Due to the low operation altitude, UAVs achieve
a very high resolution in terms of ground sampling distance and can, therefore,
compete with airborne large format digital camera system (Irschara et al, 2010).
In addition to these advantages, the UAV-images can be also used for the
high-resolution texture mapping on existing DSMs and 3D-models, as well as
for image rectification. The rectified images and derivate, like image mosaics,
maps, and drawings, can be used for image interpretation.
The implementation of GNSS systems as well as the stabilization and nav-
igation units allow precise flights with sufficient image coverage and overlap
and enabling the user to estimate the expected product accuracy preflight.
With the huge amount of divert drone advantages, we listed it in a sum-
marized way as:
1. Autonomous and stabilized: real-time capability.
2. Can fly at low altitude close to the objects where manned systems
cannot be flown (natural disaster sites, mountainous and volcanic areas, flood-
plains, earthquake and desert areas etc.…).
3. Data acquisition in cloudy and drizzly weather conditions.
4. Data acquisition with high temporal and spatial resolution.
5. More economical than human pilots.
6. Providing high-resolution texture mapping on existing DSMs and 3D-
models.
7. Real-time capability and the ability for fast data acquisition, while
transmitting the image.
8. Use in high-risk situations without endangering a human life and inac-
cessible areas.
9. Flexibility, a drone can be launched on demand.

11
 10. Timely, drones produce completely up-to-date imagery. This makes
drones suitable for monitoring projects.
11. Efficient, using a drone is fast and requires minimal staff.
12. Cost-effective, the project cost of a professional drone system is typi-
cally lower than a manned imaging aircraft.
13. Discrete, electric-powered drones make a little of noises and are rarely
disturbing people on the ground if they notice them at all.
Limitations in the use of drones
The limitations of drones compared to manned aircraft systems, first of
all, is related to the sensor payload in weight and dimension, so that often low
weight sensors like small or medium format amateur cameras are mounted on
drones, in comparison to large format cameras, drones have to acquire a higher
number of images in order to obtain the same image coverage and comparable
image resolution.
The drones payload limitations require the use of low weight navigation
units, which implies less accurate results for the orientation of the sensors.
For general mapping purposes, payload weights may vary from 200 g for
a small digital camera to 3 kg for larger digital single lens reflex or multispectral
cameras. Apart from the weight of the actual sensor and battery, a stabilizing
gimbal mount may have to be added to the payload weight. The take-off weight
of "small" UASs for mapping purposes will typically vary from around 1 kg to
5 kg.
Existing commercial software packages applied for photogrammetric data
processing are rarely set up to support drone images, as though no standardized
workflows and sensor models are being implemented (Eisenbeiss, 2009).
Based on the communication and steering unit of drones, we can state
that the operation distance depends on the range of the radio link for the rotary
and fixed wing, which is equivalent to the length of the rope for kites and bal-
loon systems used in the past.
radio frequencies may be subject to interferences caused by other systems
(remote controlled cars and model aircraft, as well as band radios), which use
the same frequencies or may suffer from signal jamming. Thus, depending on
the local situation of the area of interest.
The limitations in the drone uses are could be summarized in:
1. Limitations of the payload, the drone could not carry big weights.
2. Regulations and insurance, it is very dangerous to fly above a crowd
and to give a child a drone guidance.
3. Use of Low-cost Sensors.
4. Short flight distances (civilian use), low battery life.

12
 Chapter 3 Classification of drones

Fixed wings
A fixed-wing drone is generally composed of a central body, which hous-
es all the drone's electronics, and two wings. The aerodynamic profile of the
wings enables the drone, once in flight, to generate lift that compensates for the
weight of the aircraft.
Like an airplane, fixed-wing drones also feature ailerons, which enable
the aircraft to steer. Some drones also feature a rudder and elevators, sometimes-
even flaps.
Fixed-wing drones typically feature one engine with a propeller attached,
either a forward-mounted (tractor) propeller or a backward- facing (pusher) pro-
peller. Most propeller-powered drones include a system that folds these compo-
nents somehow especially if, an aircraft lands on its belly
(http://planner.ardupilot.com).

Fig. 3.1: Fixed wing drone, Sky Cruiser from SOUTH Company.

Fixed wings launching techniques, the listing of different techniques for

launching fixed-wing above is seen as the initial result of a concept generation.
These solutions are based on the weight of each fixed wing type.
For very heavy drones (military uses) the launcher must be a pneumatic
catapult mounted on a truck. Big size civilian fixed wings drones are launched
by catapult 4.2a, middle size drones are launched using a rubber wire, and most
small and foamy drone could be throw by hands for takeoff.
As with its launch, fixed wings presents the more complicated provision
for recovery. There are a number of solutions:
a) Skid or belly landing,
b) Guided flight into a catchment net,
c) Deployment in flight of a parachute.
Skid or belly landing is applicable only to small drones.
Guided flight into the net, this way usually comes with disastrous results
as on most occasions the drones were damaged or destroyed.
Launch equipment. This will be required for those air vehicles, which do
not have a vertical flight capability, nor have access to a runway of suitable sur-

13
face and length. This usually takes the form of a ramp along which the aircraft is
accelerated on a trolley, propelled by a system of rubber bungees, by com-
pressed air or by a rocket, until the aircraft has reached an airspeed at which it
can sustain airborne flight.
Recovery equipment. This also will usually be required for aircraft with-
out a vertical flight capability, unless they can be brought down onto the terrain,
which will allow a wheeled or skid-borne run-on landing. It usually takes the
form of a parachute, installed within the aircraft (figure 3.2), and which is de-
ployed at a suitable altitude over the landing zone (Reg Austin, 2010).

Fig. 3.2: A) catapult drone launching, b) parachute for the drone landing.

Parachute deployment, this is the most usual method; it requires the drone
to carry a parachute. One disadvantage of this method is that the parachute is at
the mercy of the wind, and its precise point of touchdown may, therefore, be un-
predictable figure 3.2b.
Rotary systems
Rotary drones, also called multi-rotors, are more complex systems than
fixed-wing.
A rotary system moves through the air by varying the power supplied to
the different propellers. This determines their revolutions per minute (RPM) and
therefore the thrust these generate.
Two key technical challenges must be met to optimize a rotary system's
performance.
First, to ensure a stable flight they need highly advanced autopilot
technology to continually choose the correct RPM for the different propellers;
making hundreds of tiny adjustments per second.
Secondly, rotary systems require very fine motor control to rapidly vary
the power sent to different motors, ensuring the propellers achieve the exact
RPM commanded by the autopilot.
In the case of a four-propeller system quadcopter model, diagonal pairs of
propellers spin in opposite directions.

14
 A quadcopter climbs and descends by simultaneously varying the RPM,
and therefore the thrust, of all four of its propellers. creates a force allowing the
drone to climb (ascend), hover or descend.
Flying forwards or backward. To pitch forwards, the RPM of the front
two motors is decreased compared to that of the back pair. With the drone
pitched forwards, the combined thrust of its four motors is also pitched
forwards, pushing the drone in that direction.
Flying sideways, or rolling, is a case of increasing the RPM of two motors
on the same side; increasing the power to the rotors on the left, while decreasing
it to the right, will 'roll'and move the aircraft to the right.
While different multi-rotors feature different numbers of propellers, these
basic flight concepts remain the same.

Table 3.1: Questionnaire for fixed wing and rotary system comparison.

Fixed wing Rotary systems

Coverage Large Small
Take-off/landing area Large Very small
Object resolution cm per pixel mm per pixel
Mechanics Relatively simple Complex
Projects Mapping Small area mapping and
inspection
Flight times and wind resistance High Low

The comparison between fixed wing and rotary drones of table 3.1 gives
an idea for the user to choose which type to use depending on project size, a
fixed wing for big projects (global) and a rotary system for small projects (lo-
cal).
The rotary system drones come with different types such as tricopter,
quadcopter, an octocopter. In this section, we will discuss the description, ad-
vantages, and disadvantages of each type.
Tricopter. From its name, we can understand that it is constituted from
three arms, each connected to one motor figure 3.3. The front of the drone tends
to be between two of the arms, the rear motor normally needs to be able to rotate
(using a normal RC servomotor).

15
 Fig. 3.3: Tricopter unmanned aerial vehicle.

Advantages: It flies more like an airplane in forwarding motion.

Disadvantages: Since the copter is not symmetric, the rear arm is more
complex since a servo needs to be mounted along the axis to assure a motor ro-
tation. Most, though not all flight controllers support this configuration.
Quadcopter. A quadcopter with four arms each connected to one motor
figure 3.4. The front of the drone tends to be between two arms like an (x), but
can also be along an arm like a (+).

Fig. 3.4: Quadcopter with four arms and motors.

Advantages: Most drones in the market are quadcopters, very simple the
arms/motors are symmetric. All flight controllers are suitable for this multirotor
design.
Disadvantages: There is no redundancy, so if there is a failure anywhere
in the system, especially a motor or propeller, the craft is likely going to crash.
Hexacopter. Six arms connected to six motors figure 3.5, the front tends
to be between two arms, or also in one arm with the form of (+).
Advantages: hexacopter can lift more payload than other kinds of copters.
If a motor fails, there is still a chance the copter can land rather than crash. All
flight controllers support hexacopters configuration.

16
 Fig. 3.5: Hexacopter with six arms and motors, Sky Walker X61 from SOUTH Company.

Disadvantages: hexacopters have more parts (motors, ESC and propellers)

then quadcopter and this makes it more expensive these additional motors and
parts add weight to the copter, so in order to get the same flight time as a quad-
copter, the battery needs to be larger (higher capacity) as well.
Y6 Hexacopter. This type of hexacopter have three arms instead of six, it
has three with six motors connected to either side of the armed figure 3.6.

Fig. 3.6: Hexacopter Y6 with three arms and six motors.

Advantages: this copter can lift more payload as compared to a quadcop-

ter and not have the same issue as a tricopter as it eliminates the gyro effect us-
ing counter-rotating propellers. In addition, in the case of a motor fail, there is
still a chance the copter can land rather than crash.
Disadvantages: Additional motors and parts additional weight, the battery
needs to be larger (higher capacity) as well.
Octocopter. Comes with eight arms, with eight motors. The front of the
UAV tends to be between two arms figure 3.7.

17
 Fig. 3.7: Octocopter eight arms and eight motors.

Advantages: More motors, more thrust, as well as increased redundancy,

can carry a big weight.
Disadvantages: More motors, higher price, and larger battery pack.
Octocopter X 8. This is a quadcopter on each arm two motors a total of
eight motors figure 3.8.

Fig. 3.8: Octocopter with four arms and eight motors.

Advantages: More motors more thrust, as well as increased redundancy.

Rather than using fewer yet more powerful motors, can carry more weight.
Disadvantages: higher price and larger battery pack.
The takeoff and landing of a copter is easier than a fixed wing because it
assures verticality.
Drones size begin from Nano to mega as the size of a car found at a varie-
ty of prices from cheap (toy) to very expensive (professional).
Choose of the drone depend on the user and from the way of use.

18
 Chapter 4 Drones structures

Unmanned Aerial Vehicles (UAV)

An unmanned aircraft system comprises a number of sub-systems that in-
clude the drone (often referred to as a UAV or unmanned air vehicle), its pay-
loads, the control station, and the remote stations figure 4.1.
A UAV, have some automatic intelligence it will be able to communicate
with its controller and to return automatically at home (point of the beginning).
UAV’s are manually guided by a radio transmitter connected through a
receiver on board of the vehicle.
A flight controller as a brain of the UAV giving the command to the Elec-
tronic Speed Control (ESC) them to motors.
In this chapter, we will know the function of each element constituting the
UAV.
The main components of a simple Unmanned Aerial Vehicle (UAV) are:
1. Frame.
2. Electronic Speed Control (ESC).
3. Motors.
4. Propellers.
5. Flight controller.
6. Batteries.
7. RC receiver.
8. RC radio transmitter.
9. Global Positioning System (GPS).
UAV airframe also includes the flight control surfaces, which could be a
combination of either aileron/elevator/rudder, or elevator/rudder, or ailerons,
airframes are made from materials such as plastic, metal or carbon fiber and
equipped with a receiver. The receiver receives information that tells it what to
do from a transmitter, receivers and transmitter have communicated using radio
frequencies (RF).
To help in flight controls a GPS provides accurate positions; GPS plays an
indispensable role in the autonomous control of UAVs because it provides an
absolute position measurement.
GPS receiver could achieve a three-meter accuracy; if more accuracy is
needed, there are differential GPS units, which could achieve centimeter-level
accuracy. The disadvantage of GPS is its vulnerability to weather factors and its
relatively low updating frequency (commonly 4Hz), which may not be enough
for flight control applications (Lafay, 2015).

19
 Fig 4.1: Main electronic elements of a quadcopter.

Drones should be powered with LiPo batteries, which are much faster
than other types of batteries because they output power faster, store a large
amount of power, and have a long life.
The battery life for a flight in good conditions (no wind or cold weather)
is for 15 minutes.
Propellers are blades attached to drones spin to create lift and move up
the drone, some drones come with special propellers that self-tighten like DJI
phantom, which does not need a key to tight.
These Propellers are attached to motors and an ESC for each controls the
speed of each motor independently and ensure the stability of the drone by
varying the speed in this case the drone is able to hover in place, climb or
descend and move in all directions.
Flying with remote control (RC) transmitter means that you will be
limited to flying line of sight and you won't have the benefit of advanced
communication that comes with smart devices like phones and tablets. Rc
transmitters have the capabilities of controlling a drone for long distances, RC
controllers do not communicate any position data or battery charge status. The
radio transitions are always present everywhere in the word, most remote control
drones use 900 MHZ for transmission and smartphones and tablet controllers
don't have the range that RC transmitter does (Lafay, 2015).
Flight controller, a flight controller is essentially a normal programmable
microcontroller, but has specific sensors onboard; a very simple flight controller
will include only a three-axis gyroscope to be able to auto level the drone.
More sophisticated flight controller includes more specific sensors like:
 Accelerometer, it measures linear acceleration in up to three axes (X,
Y, and Z). The units are normally in "gravity" (g) which is 9.81 meters per se-
cond. Accelerometers detect gravity, and it can know which direction is “down”.
This allowing multirotor aircraft to stay stable (www.robotshop.com).

20
  Gyroscope, a gyroscope measures the rate of angular change in up to
three angular axes (alpha, beta, and gamma) the units are often degrees per se-
cond.
 Inertial Measurement Unit (IMU), IMUs are electronic devices that
are capable of providing three-dimensional velocity and acceleration infor-
mation of the vehicle they are installed on at high sampling rates. They consist
of three accelerometers and three gyroscopes mounted in a set of three orthogo-
nal axes, the major problems encountered with such developed cheap IMUs, are
the need to perform the calibration process for the used sensors (Dissanayake et
al. 2001).
 Compass/Magnetometer, an electronic magnetic compass is able to
measure the earth's magnetic field and used it to determine the drone ‘s compass
direction This sensor is almost always present in the system has GPS input and
is available in one to three axes.
 Pressure/Barometer, since atmospheric pressure changes the farther
away you are from sea level, a pressure sensor can be used to give you a pretty
accurate reading for the drone’s height. Most flight controllers take input from
both the pressure sensor and GPS altitude to calculate a more accurate height
above sea level (www.robotshop.com).
 GPS, Global Positioning Systems (GPS) use satellites signals to de-
termine the specific geographic location of the drone; it registered the coordi-
nates of the takeoff points to return to it in case of emergency of automated
flights. A flight controller can have either onboard GPS or one, which is con-
nected to it via a cable. The accuracy of these GPS is not very high.
All the listed above electronic system is standard for all professional
UAV’s
The comparison of a drone with a human body is that the flight controller
is the brain, the wires are the blood vessels and nerves, and the motors are your
muscles, limbs, and hands (Issod, 2015).
Unmanned Aircraft System (UAS)
As we mentioned in the first chapter that UAV is an Unmanned Aerial
Vehicle remote control or autonomously guided. Otherwise, the UAS Un-
manned Aircraft System is very similar the UAV's one but with a ground station
control.
A UAS is a term more suitable to professional drones specialized in a spe-
cific field as military and geomatics.
The structure of a UAS comparing to a UAV is constituted from the above
listed electronic devices and sensors:
 Long distance radio transmitter (TX)/ receiver (RX) for controlling
the drone.
 Long distance First Person View (FPV) for flight monitoring.

21
  Telemetry and On Screen Display (OSD) receiving information about
the flight and display on the screen.
 Hi-Resolution cameras to move the real flight image.
 Weatherproof design.
 Accurate GPS (DJI Naza or RTK).
 Flexible payload (body of the drone, motors, etc.…).
 Autonomous flight compatible, a route tracing for long distances
flights.
Besides the electronic elements, the UAS must be equipped with:
 Field computer or tablet pc for automatic flight control.
 Workstation computer for data processing.
 Ground control points marks.
 High accuracy professional GNSS receivers.
The main parts of a typical UAS are: Autopilot, payload, communication
system and a ground control station discussed in details below.
An autopilot is a system used to guide drones without assistance from
human operators, consisting of both hardware and its supporting software. The
autopilot is the base for all the other functions of the UAS platform.
Autopilots allow defining the coordinates relative to fixed home location
or to takeoff position.
An autopilot can take control of different objectives such as:
a) Pitch attitude hold.
b) Altitude hold.
c) Speed hold.
d) Automatic take-off and landing.
e) Roll-Angle hold.
f) Turn coordination.
g) Heading hold.
The autopilot needs also to communicate with the ground station for con-
trol mode switch, to receive the broadcast from GPS satellite for position up-
dates and to send commands to UAS motors.
New autopilot systems come with new functions as follow me you can be
all time follow by your drone or flying around a waypoint with an adjusted radi-
us, etc…
GPS plays an indispensable role in the autonomous control of UAVs be-
cause it provides an absolute position measurement. A known bounded error
between GPS measurement and the real position can be guaranteed as long as
there is a valid 3-D lock. There are also differential GPS units, which could
achieve centimeter-level accuracy. The disadvantage of GPS is its vulnerability
to weather factors and its relatively low updating frequency of 4Hz, which may
not be enough for flight control applications.

22
 Payload, the payload of UAV could be a camera, or other emission devic-
es like Lidar mostly for intelligence, surveillance, and reconnaissance uses.
The type and performance of the payloads are driven by the needs of the
operational task it could vary from 200g to 4 kg (Reg Austin, 2010).
A communications system providing the data links between the Control
System and the drone through radio frequencies and it provides:
Transitioning of the flight path to be stored in the drone flight control
sytem
Transitioning real-time flight control commands.
Transmit control commands to the aircraft-mounted payloads (gimbal or
camera mounted).
Transmit updated positional information to the drone (Reg Austin, 2010).
Most UAV have more than one wireless link supported. For example, RC
link for safety pilot, WIFI link for large data sharing.
Control Stations (CS). A CS is simply the control center of a UAS sys-
tem, within which the mission is pre-planned and executed. The launching and
recovery of the aircraft may be controlled from the main CS.
It is necessary for the operators to know, on demand, where the aircraft is
at any moment in time. It may also be necessary for the aircraft to ‘know’ where
it is if autonomous flight is required of it at any time during the flight. This may
be either as part or all of a pre-programmed mission or as an emergency ‘return
to base’ capability after system degradation (Reg Austin, 2010).
From a laptop, the operator could easily input waypoints onto a map base
(google map), and providing easy access to the key and frequently used features
figure 4.2.

Fig. 4.2: Control station of a UAS system.

23
 A control station can include:
 Primary battery, used to power the LCD monitor and/or FPV glasses
and possibly the video receiver.
 Secondary battery for the transmitter.
 Mounting for the LCD monitor.
 Mounting for the video receiver
 Space for storing the RC transmitter.
 Mounting for the long-range antenna.
A laptop is not something everyone needs within the field but it can give
you some nice features like mapping, missions, telemetry, spectrum analyzer,
and disk video recorder. You will not need anything too special for computer
hardware, just as long as it is capable of running google earth at a minimum and
the battery holds for as long you need it (Glover, 2014).
Moreover, control stations are equipped with:
On Screen Display (OSD), allows the pilot to see various sensor data
sent back from the drone. One of the easier ways to include on-screen data is to
use a camera with analog output and place an on Screen display board between
the camera output and the video transmitter.
First person view (FPV), FPV describes photography where you see
what the drone is seeing. The video is beamed back to a small monitor or to a set
of special goggles, the operator is wearing. FPV gear for drones will work dif-
ferently than the purely digital cameras we are used to. These are often analog
systems and therefore use either a different second camera. This output is cou-
pled to a transmitter with its own antenna- and often needing its own battery.
The video signal is then transmitted from the drone to your ground station and
displayed on a small monitor on the side of specially designed goggles. FPV
does not require a high-resolution camera (Issod, 2015).
Smart Devices, Smartphone, and tablets can be used to display video in
real time. The difficulty with using smart devices is that most receivers are not
made to receive a video signal from a wireless video receiver. A smartphone
currently works best with the video sent via WiFi (WiFi camera) with applica-
tion to run this camera (www.robotshop.com).
With the evolution of the drone and the smart devices, tablets or phones
could replace nowadays control stations with an application installed fully func-
tioned controlling the UAS very easily and without any added equipment's, the-
se applications give you advance positioning, the first-person video controls
FPV, programmable flight routes, etc.
Smart devices during flight can display on screen drone position using
GPS, flight status, speed, battery life and flight time.
Sensor. Many of the most promising application areas for UAS relate to the
gathering of information that can be remotely sensed. This ranges from visual
range cameras gathering data for surveillance of various kinds to meteorological
instruments, to geologic surveying and crop analysis among a wide variety of
other existing and potential applications.

24
The sensors need gimbals to be mounted on UAS, a gimbal is often used to sta-
bilize a camera or a sensor, connecting a sensor directly to a UAS frame means
it is always pointing in the same direction as the frame itself.
Gimbals are high end mounting systems that reduce shake by stabilizing the
camera; some gimbals provide remote control for adjusting and rotating the
camera to capture a different perspective. Gimbals tend to reduce flight time due
to their weight (Lafay, 2015).
UAS remote sensing functions include electromagnetic spectrum sensors, gam-
ma ray sensors, biological sensors, and chemical sensors.
Cameras. In UAS applications cameras, are made useful and highly
adaptable by the addition of gimbals for pointing and stabilization software for
removing distortions caused by aircraft vibration and atmospheric buffeting.
These cameras are used as photogrammetry sensors to take aerial photography it
could be a professional camera or a sports one depending on the resolution
needed.
Infrared Detectors, a thermographic camera or infrared camera is a de-
vice that forms an image using infrared radiation, similar to a common camera
that forms an image using visible light. Instead of the 450-750-nanometer range
of the visible light camera, infrared cameras operate in wavelengths as long as
14,000 nm (14 µm).
These infrared detectors are used for the Normalized Differential Vegeta-
tion Index (NDVI) extractions.
Multispectral and Hyperspectral Sensors, Recent advances in remote
sensing and geographic information have led the way for the development of
hyperspectral sensors. Hyperspectral remote sensing, also known as imaging
spectroscopy, is a relatively new technology that is being investigated by re-
searchers and scientists in the detection and identification of minerals, and land
use (Hyperspectral Remote Sensing).
Radar, many of the most promising applications of radar-based sensing to
UAS utilize Synthetic Aperture Radar (SAR). SAR is a form of radar, which
uses relative motion between an antenna and a target region to provide distinc-
tive long-term coherent-signal variations, which are exploited to obtain a finer
spatial resolution for the production of Digital Elevation Models (DEM)
LIDAR (Light Detection and Ranging), is an optical remote sensing
technology that can measure the distance to a target by illuminating the target
with light, often using laser pulses. LIDAR technology has application in geo-
matics, archaeology, geography, geology, geomorphology, seismology, forestry,
remote sensing, and atmospheric physics (Arko Lucieer et al., 2012).
This technology helps in 3D terrain modeling by producing 3D point
clouds.
Meteorological Sensors, the use of a UAS enables the sensor to be de-
ployed to a location in the atmosphere remote from the user of the sensed data.
The National Weather Service and others have used radiosondes and operated
aircraft to reach regions of the atmosphere remote from the ground observer.
The use of radiosondes, however, is inefficient and costly.

25
 Chapter 5 Drone Software’s

An unmanned aircraft system (UAS) differ from other unmanned aerial

vehicles (UAV) by a software system of control and processing.
This software system includes a preflight software and a post-flight one,
the first one is the flight planning and the last one is the data processing for
mapping purposes.
Some of these two categories of software are open source, below we will
mention some of them and we will introduce some commercial software. From
this chapter, the reader will take an idea about the software's and their structure.
Drones flight planning software
The trend in UAS photogrammetry is moving towards the use of autono-
mous systems since manually controlled systems are highly affected by envi-
ronmental conditions.
Because of the automation of the flights and the improvements in the
workflow of data processing, UASs can be applied for practical geographical
applications. Non-experts, unlike UAVs, can control these systems.
The drone's flight planning software's deals with the drones control sys-
tems and play the role of the pilot, it can make the adjustment of the control sys-
tem, design the flight and take the guide of the vehicle. In the market, we can
find several flight planners software each of them is compatible with the control
system designed by their company manufactured below some examples of flight
planners software:
 Emotions 2 of sense fly to trace the flight line and manage it by modi-
fying the flight speed and the altitude.
 DJI Ground control for Ipads of phantom DJI.
 Q-Ground control is open source software compatible with windows.
 ArduPilot.
 Feiyu tec. DOS.
 And others…
This software’s have a very fast evolution, firstly a ground control station
in a big suitcase with cables and antennas constituting the main drone part, now-
adays a smartphone or tablet with a small application can take control of all the
flight system through a WiFi or a Bluetooth connection.
To plan a flight mission, you must go through some basic, these basics are
the same for all software with some little modifications in commands, and each
automated flight should follow these steps:
1. Running outdoor the flight path software on a pc or tablet, a window
similar to google maps will be displayed.
2. Connecting the PC to the aircraft (Bluetooth or WiFi), all the electron-
ic sensors (GPS, radio, telemetry, etc.…) of the aircraft will be connected locat-
ing the position on the screen.

26
 3. Location the area of interest and drawing the flight path by defining
the border of a polygon bounding the area.
4. Drawing a survey grid automatically inside the drawn polygon, taking
into account the overlapping side of the aerial photography by dividing the area
of interest into flight lines linked by waypoints figure 6.1. In some old software,
you can no find this function so you must draw these flight lines manually by
dropping a pin on the map of the software background.

Fig. 5.1: Tablet flight planner at Zaarour region (Lebanon) displays of Lichi Company

At any time, you can amend or cancel your planned operations. Past
flights will also be saved and can be re-used or modified for new upcoming ac-
tivity.
1. Defining the altitude of each waypoint and the angle of the mission by
rotating the survey grid, in case of fixed wings the grid must be stated with the
direction of the wind.
2. Selecting the start point which is the taking off and it is usually the
place of the aircraft and defining the landing point by a waypoint or simply se-
lecting the back home command that returns the drone to the takeoff position.
3. Adjusting the flight speed and the gimbal orientation with the number
of camera frames by second if the camera is found on board the drone.
4. Camera adjustment to take simultaneous photos in case of DSLR
cameras (time lapse), in case of the camera is built in adjustment is made direct-
ly from the software application with the mission planning.
The mission planning software displays the GPS number of satellites
tracked, battery life and flight duration.
Figure 5.1 shows the waypoints dropped manually forming the flight path,
the position of the aircraft expressed by latitude and longitude with the begin-
ning and the end of the mission.
We define three scenarios in the flight planning module:
 Documentation of a surface with the flat or moderate terrain.

27
  Exposure of a rough/ mountainous terrain like hazard areas.
 3D modeling of buildings and other objects.
Some new tablet personal computers applications with new functions like
flips and rolls (Abbeel et al., 2007), collision avoidance (Bellingham et al., 2003
and Pettersson and Doherty, 2004), automated target tracking (Nordberg et al.,
2002) and operations like the “follow me” modus.
Nowadays, waypoint navigation for UAVs is a standard tool (Niranjan et
al., 2007). Thus, the autonomous flight based on defined points in a global coor-
dinate system is possible for most UAV systems. For the autonomous flights of
UAVs, a start and a home point have to be defined relative to its coordinate.
Moreover, some packages allow in addition to the waypoints, the definition of
lines, paths, boundaries and no-go areas (Gonzalez et al., 2006; Wzorek et al.,
2006).
Since most of the autonomous systems are stabilized, it can be expected
that the flight trajectory of the autonomous systems is more stable. Thus, for the
automation of the workflow, the orientation values coming from the navigation
units can be used as an approximation for the image orientation and for the fast
production of overview images (Eisenbeiss, 2009).
Drones mapping software
After finishing the flight mission successfully, a post-flight image pro-
cessing is needed to extract data, this part is the photogrammetry episode de-
scribed in detail in chapter 6.
For data processing, a powerful PC station and a professional software
can make the job, we listed below a series of open source software’s found on
the net it could help students and researchers to do their job:
Airphoto SE offers essential features needed for rectification of oblique
aerial imagery with geo-referencing. It is capable to make an automatic correc-
tion for radial lens distortion. It is designed for beginners or experienced users
for combining aerial images with maps, orthophotos, and satellite images
(http://www.uni-koeln.de/~al001/airphotose.html).
Fiji, Fiji is an image-processing package based on Java3D and many
plugins organized into a coherent menu structure (http://fiji.sc/Fiji).
ImageJ is a public domain Java image-processing program for Macin-
tosh. It can display, edit, analyze, process, save and print 8-bit, 16-bit and 32-bit
images(http://imagej.nih.gov/ij/index.html).
MapKnitter, MapKnitter is a free and open source tool for combining
and positioning images in geographic space into a composite image map.
Known as “orthorectification” or “georectification” to geographers
(http://mapknitter.org/).
Visual SFM, VisualSFM is an application for 3D reconstruction using
structure from motion (SFM). The reconstruction system integrates several other
projects: Bundle Adjustment, and Linear-time Incremental Structure from Mo-
tion (http://ccwu.me/vsfm/).

28
 CMPMVS, CMPMVS is a multi-view reconstruction software. From im-
ages to texture mesh (http://flightriot.com/post-processing-software/cmpmvs/).
CloudCompare, Cloud Compare is a 3D point cloud (and triangular
mesh) processing software. It was also meant to deal with huge point clouds
(typically more than 10 million points, and up to 120 million with 2 Gb of
memory)(http://www.cloudcompare.org/).
Meshlab, MeshLab is an open source, portable, and extensible system for
the processing and editing of unstructured 3D triangular meshes. The system is
aimed to help the processing of models arising in 3D scanning, providing a set
of tools for editing, cleaning, healing, inspecting, rendering and converting
meshes (http://meshlab.sourceforge.net/).
Drone image processing is based on a classical photogrammetric model,
which is enhanced in turn by a powerful computer vision algorithm. This ena-
bles the automatic extraction of numerous key points in the images and optimiz-
es camera parameters such as external orientation and camera model.
Other quality processing programs can alternatively be used, such as Pho-
toScan by Agisoft, Pix4D, SkyPhoto of SOUTH Company and many others.
All these softwares have a first processing phase employed by identifies
and extracts matching key points in the overlapping sections of the aerial images
acquired by the drone. These key points are entered into an equation that deter-
mines the precise position and orientation of each image, as well as the internal
and external camera parameters. The next step of the process is point cloud den-
sification, which is required to obtain a highly accurate 3D model. This is used
to generate a digital surface model (DSM) and orthophotos.
Obviously, we won´t tell you which software is “the best”, but we can
definitely present you some of the most interesting solutions which we have
tested. As a good mapping software, we can only recommend SkyPhoto from
SOUTH surveying company, if you need to generate high-resolution georefer-
enced orthophotos or extraordinarily detailed DEMs. Which we will speak about
its structure in details. This software creates a 3D model from multiple digital
photos of the area to map. Moreover, drones use airborne GPS data in order to
fulfill the geo-referencing task. If you need a better accuracy, into the 3D model
can be imported and matched GCP (Ground Control Points) or use a drone with
Real Time Kinematic (RTK) mounted on board which can give you directly
georeferenced data.
The 3D mapping program I would like to present you in detailed is SKY
PHOTO from SOUTH surveying company, which we are grateful to give us the
permissions and material to introduce you the structures of drone processing
software. With this solution, you will be able to handle large volumes of data
from drone orthophotos by bringing them into a virtual environment.
This software is linking between drone’s data and traditional geodetic
surveying; it gives you a chance to explore your project with your UAS with a
very high accuracy.

29
 All UAV software for data processing has the same interface principle,
the workflow is very similar with some advanced function, a basic notion ad-
vanced way is as follows figure 5.2.

Fig. 5.2: Basic drone data processing steps.

After getting photos from the camera the first step in a data processing
software begins with the picture insertion, a structure from motion method align
the picture to each other see chapter 7, the bundle adjustment of these photos
allow the generation of 3D point clouds. A mesh of triangulated irregular net-
work (TIN) interpolated from point clouds for a Digital Surface Model extrac-
tion (DSM), some software has the possibility to add Ground Control Points
(GCP) for the georeferencing.
All the photos joined to form a mosaic of the whole flight area draped on
the DSM to produce an orthophotoplan.
Some more advanced photogrammetric software’s own more functions
than a basic software helpful for specialists in the field of geomatics, as an ex-
ample of SkyPhoto.
Processing workflow
All professional drones data processing software designed for transform-
ing low-altitude aerial images into consistent and accurate points cloud, DEM
(Digitized Elevation Modeling), DOM (Digitized Orthophoto Modeling) mosa-
ics, etc. These software's features sharply are not only one-key processing for
workflow automation but also advanced settings and editable output options.
The special functions of this software, begin from indoor camera calibra-
tion, dodging process, accuracy quality report, measurement tool, 3D modeling
generation and browse, DLG (Digitized Line Graphics) production based on ste-
reo image pair and so on, the sequential processing actions are summarized in
figure 5.3.

Fig. 5.3: SkyPhoto UAS processing data.

In figure 5.3 an example of SkyPhoto software data processing path with

the same principle of advanced professional functions in red fonts as Importing

30
GPS and IMU data of each photo these data obtained from the flight control
software of the drone as a text file figure 5.4.

Fig. 5.4: GPS/IMU file from flight control of a SOUTH drone.

This file in figure 5.4 containing the name of the image, coordinates,
flight altitude and IMU data Heading, Pitch and Roll.
These three rotations are the transformation between the image reference
system, and a flat, projected mapping plane, most often in Universal Transverse
Mercator (UTM).
Omega – Rotation about the X-axis.
Phi – Rotation about the Y-axis.
Kappa – Rotation about the Z-axis.
These angles will produce the same result as all of the above-described
transformations (based on Heading, Pitch, Roll), mapping the raw image directly
into the UTM mapping plane figure 5.5.

Fig. 5.5: Rotation of the aerial photography.

The drone in on the flight path taking aerial photography the GPS get the
coordinates of the middle of the photo and the IMU calculate its rotation in a
Universal Transverse Mercator (UTM) system.

31
 In Skyphoto the aero-triangulation is made basing to GPS/IMU data and
as we said that the GPS accuracy is not very high about approximately 2 meters,
we must refer to Ground control points (GCP) surveyed by accurate geodetic
instruments.
When these previously surveyed points, GCP's added in the software a re-
aero-triangulation is needed to reprocess the data after matching the GCPs.
This SkyPhoto aero-triangulation editing function allows the user to con-
trol the accuracy and adjust it if possible based on a plotted report of errors.
The intellectualized aerial triangulation algorithm satisfactorily deals with
tough cases like images from unstable flight attitude (Kappa or Omega angle out
of tolerance) and sparse textures, for example, deserts, a large water area with
just a little land, etc. The overlap percentage and rotating angle of images are
very little restricted.
High-precision POS data from airborne GNSS-RTK system successfully
gets you to minimize the huge efforts in dealing with ground control points
fieldwork, and you may go straight to adjustment then mapping without the
GCP concern.
Similar to all other photogrammetry software’s SkyPhoto can generate a
mosaic, a DSM and orthophotos, more than that SkyPhoto has the possibility to
DEM edit, this function transforms the DSM to a DTM by subtraction of the
trees and buildings heights.
Millions of orientation points are attributed and even reach to hundreds of
millions after densification. Instead of monochromatic points cloud, the colorful
output is more convenient for users to view and analyze the shape and properties
of surface features. In addition, you may browse the points cloud in the software
like the way that you do with a 3D laser scanner.
Many terrain analysis software based on DEM can extract contour lines,
ridges, and thalwegs, after aero-triangulations and DEM edit SkyPhoto allow the
user the quick extraction of terrain features in DLG format without any addi-
tional software.
UAS use a simple digital camera, these cameras are not metric to calculate
the cameras internal parameters that give us the possibility to use it as metric
one, SkyPhoto has a specialized camera calibration program is included for im-
age distortion correction, as similarly required by all other professional aerial
photogrammetry software solutions. By manually inputting camera parameters
(eg.principal point, principal distance, pixel size, etc.), you may easily finish the
procedures indoors with a desktop LCD monitor. It is to either be done before or
after photography, but normally, better before the flights
(www.southinstrument.com).
As an example of photogrammetry processing figure, 5.6 shows the work-
flow of SkyPhoto software by South surveying instruments company china.
This workflow is similar to the workflow of the aerial mapping system
with some different, it consists of the definition of the new project, importing

32
the aerial images and their position file for processing and image stitching inser-
tion of GCP’s in case of non-accurate GPS image position files for the genera-
tion of dense point clouds, Digital Surface Model (DSM) and Digital Ortho
Model (DOM).

Fig. 5.6: SkyPhoto photogrammetry processing workflow.

33
 Chapter 6 Drones photogrammetry methods

Traditional photogrammetry methods begin from analogical, analytical to

digital photogrammetry lead to a new drones photogrammetry methods more
simple based on fully computerized systems (software), these software open a
new era in aerial and terrestrial photogrammetry.
Type of photogrammetry
There are two main types of photogrammetry that are used today in geo-
matics, which are close-range and aerial photogrammetry. As the name im-
plies, aerial photogrammetry is by attaching downward facing cameras to an air-
craft and taking images from above. According to Matthews (2008), Aerial pho-
togrammetry utilizes large-format imagery and ground control points (GCP) to
georeferenced an earth portion in a virtual model. This virtual model allows us
to take horizontal and vertical measurements.
Aerial photogrammetry is commonly used for mapping surveys, digital
terrain models (DTM), and digital orthophotos (Matthews, 2008).
The other type of photogrammetry is close-range, which is very similar to
aerial photogrammetry as many of the basic principles. The main difference be-
tween the two is the distance from the camera to the object. In close-range pho-
togrammetry, the object-to-camera distance is less than 300 m (Matthews 2008).
A variety of different camera configurations and platforms are included under
close-range, including low-level aerial for light sports aircraft and helium-filled
blimps (Matthews 2008). Also included in close-range photogrammetry terres-
trial and UAVs systems.
Close-range Photogrammetry has not always been a preferred method un-
til the appearance of the new structure from motion software.
Another difference from aerial photogrammetry is the ability to use sim-
ple non-special cameras due to the advancements in 3D modeling software and
digital camera resolution.
Close range photogrammetry (CRP)
The same basic principles of traditional aerial photogrammetry can be ap-
plied to pictures taken from lower altitudes or from the ground. Terrestrial,
ground-based, and close-range photogrammetry (CRP) are all descriptive terms
that refer to photos taken with an object-to-camera distance of less than 300 m,
the case of drones photogrammetry. Since the same basic principles of a terres-
trial and suspended from a light sports aircraft (low-level aerial), both types of
nontraditional photogrammetry are referred to in close-range photogrammetry
(Matthews, 2008).
The reduction of a three-dimensional object to a two-dimensional image
implies a loss of information such as the object areas which are not visible in the
image that cannot be reconstructed from it.
For the reconstruction of an object from photographs or images, it is,
therefore, necessary to describe all elements which contribute to this process,

34
such as light sources, properties of the surface of the object, the medium through
which the light travels, sensor and camera technology, image processing.
Close range photogrammetry has significant links with aspects of graphics
and photographic science, for example, computer graphics and computer vision,
digital image processing, computer-aided design (CAD), geographic information
systems (GIS) and cartography. Traditionally, there are also strong associations of
close-range photogrammetry with the techniques of surveying, particularly in
the areas of adjustment methods and engineering surveying.
The application of CRP is very wide we list below some specialist area of
uses such as:
 Architectural photogrammetry: architecture, heritage conservation, ar-
chaeology.
 Engineering photogrammetry: general engineering (construction) ap-
plications.
 Industrial photogrammetry: industrial (manufacturing) applications.
 Forensic photogrammetry: applications to diverse legal problems.
 Biostereometrics: medical applications.
 Motography: recording moving target tracks.
 Multi-media photogrammetry: recording through media of different
refractive indices.
 Shape from stereo: stereo image processing (computer vision).
A very important question to ask, how does automated CRP work?
1. Data collection: Multiple overlapping photos from different locations
2. Automated feature matching: Over 2000 matches and nearly 1000 in-
correct matches.
3. Derived interpolated surface geometry (mesh).
Figure 6.1 shows the close-range photogrammetry sequential processing
workflow.
Beginning with the image preparation which could be photo enhancement
contrast illumination and selection, Keypoint Detection the selection of similar
pixels between photos, Keypoint Matching for the two first images than to the
remaining ones, Loop Closing image preparation for the Bundle Adjustment step
of least square calculation which lead to the Transformation to absolute coordi-
nates. After that comes the Image Rectification based to the absolute coordi-
nates for the Dense Depth Estimation to build a 3D model by Triangle Mesh
Generation of the Model Fusion.

35
 Fig. 6.1: CRP processing workflow.

As for traditional photogrammetry works, in the CRP work some things to

avoid such as:
 Very dark surfaces.
 Reflective surfaces.
 Transparent surfaces (including water).
 Uniform textures and solid color surfaces.
 Moving light sources/shadows.
 Capturing your own shadow.
In the last years the evolution of the aerial photogrammetry was very fast
and lead to new advanced computer technologies based on very powerful soft-
ware, the difference between traditional aerial photogrammetry and close range
photogrammetry in table 6.1 was made by Andrew Marshall in the school of
surveying of the University of New South Wales in 1989, in this year as all we
know that the drone wasn't used for civilian photogrammetry but the close range
photogrammetry still has the same principle and procedure the only change is
advanced in technology.

36
 Table 6.1: The difference between aerial photogrammetry and close-range photogrammetry
(Marshall, 1989).
Aerial photogrammetry Close range photogrammetry
Relief is small comparing to flying height. Objects may have truly spatial characteristics
(large depth).
Precision requirement differs for height and Precision in all three coordinates may be
planimetry. equally important.
The entire format is usable. A restricted format is likely.
Vertical photography is used exclusively. Spatial nature of the object necessitates pho-
tography with varying position and orienta-
tion.
Target only for control points if at all. Maybe possible to target all points.
A large block may consist of thousands of The total number of photography is usually
photographs. too small.
Auxiliary data have only limited accuracy. It is possible to determine camera parameters
accurately.
A fairly standardized approach for all appli- Flexible approach required due to differences
cations. from project to project.

Table 6.1 of A.R. Marshall (1989) still valid and could be applied as a
comparison between drone photogrammetry and plane photogrammetry with
some modifications such digital cameras precision, high planimetry accuracy
GPS one, point cloud extractions, 3D models, etc…
In CRP camera precision plays a very important role, especially in the
spatial precision of the output data.
A few years ago metric cameras were used in CRP, "A metric camera is a
general term applied to a camera which has been designed for surveying and has
a well-defined inner orientation. That is a camera with a good lens of small dis-
tortion, in which the position of the principal point can be located in the image
plane by reference to fiducial marks. All Cameras not possessing these charac-
teristics can be defined as simple or non-metric Cameras" (Adams, 1980).
It is around this time (1980) that non-metric cameras were established as a
suitable tool for close-range photogrammetry, and that the accuracy of projects
using non-metric cameras could equal those using metric cameras (Karara, and
Faig, 1980).
Nowadays non-metric cameras could be used in close range photogram-
metry after camera calibration procedure for obtaining the internal parameter
and characteristics of the camera.
In his paper, Xie Feifei spoke about The Design of Four-combined Wide-
angle Camera.
There are a number of developed products of the four-combined wide-
angle camera (Xie Feifei et al., 2014).
The structure of the new four-combined wide-angle camera is shown in
figure 6.2.

37
 Fig. 6.2: Combined oblique cameras system.

These oblique aerial images are used in photogrammetry to create three-

dimensional models because they capture all sides of an object due to the differ-
ing angles captured by the various cameras figure 6.2. This type of imagery is
used most commonly for three-dimensional urban mapping (Redweik, 2012).

Distortions and Camera calibration

The model describing the geometric properties of the camera and lens sys-
tem is known as the inner or interior orientation, sometimes also referred to as
camera intrinsic (Luhmann et al., 2006).
As stated in Collinearity, the 2D image coordinate system has its origin at
the center of the image. The principal point is the orthogonal projection of the
projection center on the sensor, and is not necessarily the same as the center of
the image, hence the necessary computation of principal point offset values.
The interior orientation describes the parameters required to allow the
principles of collinearity to be applied to distorted images. The exterior orienta-
tion (also referred to as camera extrinsic) describes the position (for example x,
y, z) and attitude (roll, pitch, and yaw, or omega, phi, and kappa) of the camera’s
projection center when the image was taken. The SfM approach, used, for ex-
ample, by Bundler, Agisoft Photoscan, Photomodeler and Pix4D mapper soft-
ware performs an automatic calibration using, in the case of digital images, some
of the exchangeable image file format (EXIF) metadata in the image file as a
starting point.
This defines the camera’s interior orientation and simultaneously calcu-
lates the exterior orientations using tie points identified on the input images in a
process known as bundle adjustment (a reference to the bundles of light rays
converging on the optical center of each camera).
One time internal parameters of a camera are known, any point in the
space can be fixed by the intersection of two rays of light that are projected from
two different photos. This is the principle photogrammetry to align pictures and
produce models. However, there are two major factors that can affect the accu-
racy of the resulting model.
1 – The first is a system error due to lens distortion, this error causes the
point to shift from its true position to a skewed position (Dai and Lu, 2010).

38
 This is due to a combination of decentering distortion and radial distor-
tion, decentering distortion is caused by the camera lenses not being perfectly
centered in relation to each other, and radial distortion is a distortion in each
lens. This type of error is easily corrected by the photogrammetry software (Dai
and Lu, 2010).
2 – The second type of errors is errors due to human factors. Most human
errors are attributed to the imprecise marking of points in two different photos.
This can be overcome by marking the points in three or more photos This is why
most programs require the points to be marked in at least three photos (Dai and
Lu, 2010).
To reduce these distortions, we need a camera calibration, camera calibra-
tion is the determination of principal distance (c), principal point coordinates
(Xp and Yp) and lens distortion parameters (K1, K2, K3, P1, and P2).
Without calibration, high accuracy cannot be achieved, for sub-millimeter
surveying accuracy, the camera calibration should be carried out in laboratories.
The photogrammetric approach in early vision calibration methods, where
an object with known characteristics is used to infer the internal and external
parameters from images. The advent of structure from motion and projective
reconstruction without knowledge of camera parameters has made this unneces-
sary; scene reconstruction is attempted where no calibration objects are present.
In briefing we discussed four categories of camera calibration approaches:
1. Classical calibration: using the geometry of known objects in the world
(Leibowitz, and College, 2001).
2. Auto-calibration assuming fixed internal parameters: assuming that the
same or an identical
camera is used for each view of a scene.
3. Auto-calibration with varying internal parameters: relaxing (partially)
the assumption of identical cameras.
4. Special or constrained camera motion: solving the special cases where
the motion between cameras are known to be of a particular form (Leibowitz,
and College, 2001).
Classical calibration using known objects originated with photogramme-
try, aimed at extremely accurate camera calibration and making use of specially
designed calibration devices. The Direct Linear Transformation (DLT) of Ab-
del-Aziz attempts to apply the calibration from objects to common types of
cameras in a computationally direct manner (Abdel-Aziz, Karara, 1971).
The calibration grids of Tsai represent a move towards automated calibra-
tion from images. Tsai uses a planar calibration pattern (which has come to be
known as the Tsai grid) consisting of a grid of black and white squares. An im-
age of grid points with known world coordinates allows calibration from a single
view by computing the projection matrix from the images of points and their
known world coordinates. The calibration method of Tsai is the most applied
calibration method in most of photogrammetry software (Tsai, 1986).

39
 Auto-calibration was first explored by Faugeras et al, using the Kruppa
equations. The Kruppa equations embody the constraint that the epipolar planes
tangent to the absolute conic project to corresponding epipolar lines tangent to
the Image of the Absolute Conic (IAC) in each view (Faugeras
et al., 1992).
Auto-calibration with varying internal parameters, the constant inter-
nal parameter of many auto-calibration techniques is impractical. The assump-
tion that a single camera with identical parameter used for the entire sequence is
often invalid, although some of the parameters may remain constant, the focal
length and principal point change. This is the approach taken by Heyden and
Astrom allowing for the varying, unknown focal length and the principal point
of square pixel cameras. An iterative minimization with the square pixel para-
metrization of the camera is used to compute metric rectified cameras (Heyden,
Astrom,1997).
Special or constrained camera motion auto-calibration, Certain cam-
era motions, reduce the number of ambiguities in a calibration problem or allow
a simple solution. The cases mentioned here include pure translation or rotation
of cameras.
When a camera is rotating about its optic center and not translating, all
points in a pair of images are related by a homography. Furthermore, this
homography is the infinite homography between the cameras (Leibowitz, and
College, 2001).
With the advance of the software technologies camera calibration is made
automatically inside the processing workflow. To meet accuracy requirements of
three-dimensional (3-D) photogrammetric mapping, the conventional photo-
grammetric aircraft flying mission is usually exactly designed for some flying
parameters, such as overlap, side-lap, flying height, the direction of flight, veloc-
ity of flight, etc. For the UAV-based photogrammetric mapping system using a
high-resolution non-metric CCD camera, camera calibration has to be performed
to provide the interior orientation parameters. A listing of 91 articles on aspects
of optical camera calibration for the period 1889 until 1951 is provided by (Roe-
lofs, 1951). A large number of camera calibration approaches have been pro-
posed since 1951.
Camera calibration is the main feature in photogrammetric 3D object res-
titution. Calibration parameters such as principal distance, principal point, and
lens distortion are usually determined by a self-calibrating bundle adjustment
based on the collinearity equations and additional correction functions.
The self-calibrating bundle adjustment is a very flexible and powerful tool
for camera calibration and systematic error compensation, and it provides for
accurate sensor orientation and object reconstruction while treating all the sys-
tem unknowns as stochastic variables (Hongxia et al., 2007).

40
 Bundle adjustment
Bundle Adjustment refines a visual reconstruction to produce jointly op-
timal 3D structure and viewing parameters. ‘bundle’ refers to the bundle of light
rays leaving each 3D feature and converging on each camera center figure 6.3.
Bundle adjustment is a mathematical model, which allows the determina-
tion of camera position, camera orientation, object point coordinates and camera
calibration parameters (SMITH and PARK 2000) (see Figure 6.3 below).

Fig 6.3: Typical camera configuration of bundle adjustment in close-range photogrammetry

The bundle adjustment method of figure 6.2 is summarized graphically,

the first image i, joined by corresponding feature points to image i+1 and so on
for other images, the process continues for building the 3D model and trans-
forming all corresponding points into 3D point clouds.
The bundle adjustment image matching process relies on a least squares
solution with geometric constraints, where matching image points are deter-
mined together with the object space coordinates of the corresponding object
points figure 6.2. Bundle adjustment is really just a large sparse geometric pa-
rameter estimation problem, the parameters being the combined 3D feature co-
ordinates, camera, and orientation parameters. The basic model for photogram-
metric bundle adjustment is based on the well-known Collinearity equations
(Mikhail et al., 2001; McGlone, 1989):

With:
O = center of the projection,
f = calibrated focal length,

41
 a = coefficient of the rotation matrix.
Many scientists worked on bundle adjustment methods, all of these meth-
ods are based on the same Principe and structure with some modification, these
methods are Gradient Descent Method, Newton-Rhapson Method, Gauss-
Newton Method, and Levenberg – Marquardt Method. These four are the most
popular techniques for nonlinear least squares optimization, the last two are used
to optimize nonlinear least squares in particular.
Gradient Descent Method:
 Based on the first-order optimization algorithm.
 To find a local minimum of a function using gradient descent, one
takes steps proportional to the negative of the gradient of the function at the cur-
rent point.
 It is robust when x is far from optimum but has poor final conver-
gence.
Newton – Rhapson Method:
 Based on the second order optimization method
 Newton's method can often converge remarkably quickly, especially if
the iteration begins "sufficiently near" the desired root.
 For a quadratic function, it converges in one iteration.
 For other general function, its asymptotic convergence is quadratic.
 The disadvantage of this method is the high computation complexity
of H .
Gauss-Newton Method:
 The Gauss-Newton algorithm is a method used to solve non-linear
least squares problems.
 For well-parametrized bundle problems under an outlier-free least
squares cost model evaluated near the cost minimum, the Gauss-Newton ap-
proximation is usually very accurate.
Levenberg – Marquardt Algorithm:
 This method interpolates between the Gauss-Newton algorithm and
the method of gradient descent.
 When far from the minimum it acts as the steepest descent and it per-
forms gauss newton iteration when near to the solution.
 It takes into account the best of both gradient descent and gauss new-
ton method.
General Facts about optimization methods:
 Second order optimization methods like Gauss-Newton and LM re-
quires a few but heavy iterations.
 First order optimization methods like Gradient descent requires a lot
of light iterations.
Bundle adjustment is really just a large sparse geometric parameter esti-
mation problem, the parameters being the combined 3D feature coordinates,

42
camera poses, and calibrations. bundle adjustment and similar adjustment com-
putations are formulated as nonlinear least squares problems (Brown, 1976;
Cooper, Cross,1991; Atkinson, 1996).
Bundle block adjustment
The name bundle block adjustment is based on the fact that the rays from
the projection center to the photo points are building a bundle of rays, the bundle
block adjustment is using the photo coordinates, based on these photo coordi-
nates the bundle block adjustment is leading to more accurate results than the
other methods. Of course, some additional corrections like self-calibration with
additional parameters are improving the results as well as additional observa-
tions like GPS- positions of the projection centers.
The Bundle block adjustment provides three primary functions:
1. It determines the position and orientation of each image as exterior ori-
entation parameters. In order to estimate the exterior orientation parameters, a
minimum of three/four GCPs is required for the entire block, regardless of how
many images are contained within the project.
2. It determines tie points ground coordinates on the overlap areas of mul-
tiple images. The highly precise ground point determination of tie points is use-
ful for generating control points from imagery.
3. It minimizes and distributes the errors associated with the imagery, im-
age measurements, and GCPs. The bundle block adjustment processes use statis-
tical techniques to automatically identify, distribute, and remove error (Rüther et
al., 2012).
Structure from Motion (SFM)
Structure from Motion (SfM), was introduced that allows the extraction of
the 3D structure of an object by analyzing motion signals over time (Dellaert et
al., 2000). The SfM technique can be applied to large collections of overlapping
photographs to obtain sparse point clouds for a wide range of objects, such as
buildings and sculptures (Snavely et al 2007; Snavely et al 2006). The power of
this technique was demonstrated by (Snavely et al 2007) who developed the
Bundler software and used it to construct 3D models.
The term Structure-from-Motion has evolved from the machine vision
community, specifically for tracking points across sequences of images occupied
from different positions (Spetsakis and Aloimonos, 1991; Boufama et al., 1993;
Szeliski, Kang, 1994) (figure 6.4).
Structure-from-Motion (SfM) operates under the same basic tenets as ste-
reoscopic photogrammetry, the 3D structure can be resolved from a series of
overlapping images. However, it differs from traditional photogrammetry by
camera positions and orientation that are solved automatically without the need
to specify the network of targets with known X, Y, and Z.

43
Fig. 6.4: Structure-from-Motion (SfM) functional schema requires multiple, overlapping pho-
tographs of the object.

Many SFM software's are found as a commercial and open source espe-
cially for terrestrial and aerial Unmanned aerial vehicles (UAV) uses, open
sources SfM packages have very limited post-processing functions unlike com-
mercial software for a variety of tools and models, due to their high prices and
fast evolution.
SfM involves the 3D location is determined through automatic iden-
tification of matching features in multiple images, these features called tie
points, used to establish both interior and exterior orientation parameters of
cameras.
In photogrammetry, 3D geometry is obtained by creating images of the
same object from different positions. This makes a single point on the object
visible as a pixel in multiple images. For each image, a straight line can be
drawn from the camera center through the pixel in the image. These lines will
intersect at one point (tie point), which is the 3D location of the object point.
These tie points are tracked from image to image pixel by pixel estimating
camera positions and object coordinates which are then refined iteratively using
non-linear least-squares (Snavely et al., 2008). SfM method identifies features in
each image that is invariant in scale, rotation and in illumination changes condi-
tions and 3D camera viewpoint (Lowe, 2004).
(Micheletti et al., 2014) demonstrated that a big the number of images
produces a very high spatial resolution data at a good accuracy because of the
big overlapping areas covered by pictures. Transparent, reflective or homogene-
ous surfaces present difficulties because incorrect features can be linked during
the automatic feature-matching process (Autodesk, 2014). Tie points number
depends on image texture and resolution; they are automatically identified over
all scales and locations in each image.

44
 Sufficient tie points allow for the reconstruction of the relative position of
all images. Additionally, known points or ground control points(GCPs) with 3D
world coordinates should be added to obtain scale and absolute coordinates.
The SfM part of the process generates a sparse point cloud comprising tie
points identified and matched across the input images. In order to construct the
sparse point cloud, several steps are involved such as feature extraction, feature
matching, and bundle adjustment. The SfM algorithm needs to estimate the inte-
rior and exterior orientations for each image by combining all the relative orien-
tations of the image pairs in the form of their fundamental matrices (Verhoeven
et al 2013). Once complete, a technique called image triangulation is used to
calculate the relative position and orientation for each image in every pair. The
overlapping pairs are then combined to form a single block, achieved by a bun-
dle adjustment, because it necessitates adjusting the bundles of rays between
each camera’s projection center and the set of projected 3D points until there is
minimal discrepancy between the positions of the observed and re-projected
points (the image distance between the initial estimated position of a point and
its ‘true’ or measured value) (Verhoeven et al., 2013).
A bundle adjustment step comes after the key point's detection to find 3D
point positions and camera parameters that minimize the re-projection error,
bundle adjustment, and similar adjustment computations are formulated as non-
linear least squares problems (Cooper, Cross, 1988; Granshaw, 1980; Atkinson,
1996; Karara, 1989; Wolf, Ghilani, 1997). Least square problems are the differ-
ences between the observed feature location and the projection of the corre-
sponding 3D point on the image plane of the camera, the problem can be cast as
a re-weighted non-linear least squares problem (Hartle, Zisserman, 2003).
Bundle adjustment is really just a large sparse geometric parameter esti-
mation problem, the parameters being the combined 3D feature coordinates,
camera poses, and calibrations.
After bundle adjustment in data sets, processing comes the point clouds
generations in a relative ‘image-space’ coordinate system, which must be
georeferenced to a real world using a 3D transformation by adding a small num-
ber of known control points (Doumit, Kiselev, 2016).
The technique is based on identifying matching features in images that are
taken from different viewpoints. Image features are identified by the scale invar-
iant feature transform(SIFT) algorithm (Lowe, 2004), which is robust in terms
of its feature descriptors for image features at different viewing angles. Based on
these SIFT matches, the camera positions, orientations, radial lens distortion,
and finally the 3D coordinates of each SIFT feature are calculated using a bun-
dle block adjustment. The 3D positions of the SIFT feature essentially form a
3D point cloud that captures the structure of an object.
The point cloud is known as a sparse point cloud that can be densified
with a more recent technique called multi-view stereopsis (Furukawa, Ponce,
2009). The stereopsis algorithm takes the output from the Bundler algorithm,

45
camera positions, orientations, and radial undistorted images, and applies a
match, expand, and filter procedure. It starts with the sparse set of matched key
points, and repeatedly expands these points to neighboring pixel correspondenc-
es, and finally applies visibility constraints to filter out false matches (Furukawa
& Ponce, 2009). The algorithm is implemented in the Patch View Multi-Stereo
(PMVS2) software tool. SIFT, Bundler, and PMVS2 work in sequence to gener-
ate an extremely dense 3D point cloud just from overlapping photographs. The
PMVS2 point cloud contains the following information for each point: XYZ co-
ordinates, point normal (i.e. the direction of the slope through the point), and
point RGB color values (i.e. derived from the photographs. The dense point
cloud can then be used as the basis of a triangulated irregular network (TIN) or
mesh, onto which textures generated from the input images can be projected.
SFM requires the uploads of multiple images of an object and the results
are produced fully automatically without user interaction. The principal disad-
vantages of SFM are that the mathematical process is not transparent and many
research has been done and published without being confident about metric ac-
curacy and reliability of their results.
Aerotriangulation
An important and critical phase in photogrammetric mapping is rectifying
the aerial images to their appropriate place on the surface of the earth. This is
accomplished by collecting horizontal and vertical data, to ascertain the spatial
location of a number of features that are visible and measurable on the aerial
images.
Geometric stability requires a minimum of four points of known positions
spaced in the corners of a full stereo model be used to fully rectify On a project
involving a few stereo models this may be a conventional ground surveying en-
terprise. The expense and time required to collect the ground survey data in this
manner may render the mapping project impractical (Falkner, Morgan, 2002).
Computer processing has played a major role in driving mapping scien-
tists to develop rigorous and efficient mathematical protocols that allow for the
densification of stereo model control from a minimal number of strategically
positioned ground survey points. This procedure is generally referred to as aero-
triangulation. Analytical software available today, with its built-in quality
checks, has made aerotriangulation the preferred method of image adjustment to
the earth for photogrammetric mapping.
In this book will not discuss the theory of these processes, but rather it
will give the reader a necessary guidance and explanation of procedures to plan
and estimate the efforts required to perform satisfactory aerotriangulation for a
photogrammetric mapping project (Falkner, Morgan, 2002).
Aerotriangulation Computer Processing
After the collection of the field control point, they are imported into the
computer and processed through an aerotriangulation software module proce-
dure including:

46
  first, each stereo model is processed through a relative orientation rou-
tine involving a least squares adjustment of collinearity equations. This solution
produces individual model coordinates unrelated to any reference system.
 Second, a strip formation procedure joins the independent stereo mod-
els through a three- dimensional transformation (X, Y, Z). A series of equations
link successive models by common pass points. The coordinates, at this stage,
remain in an arbitrary reference scheme.
 Then each strip undergoes a polynomial adjustment which produces
preliminary ground coordinates for all of the photo control points, notes that the
control points are in a coordinate system of the project.
 A simultaneous bundle adjustment provides a fully analytical aerotri-
angulation solution and the entire block of data passes through an iterative
weighted least squares adjustment until a convergent “best fit” solution is ob-
tained. An RMSE error is noted so that the observer can judge how far the coor-
dinates of each point were mathematically “stretched” out of position in order to
resolve a solution (Falkner, Morgan, 2002).

47
 Chapter 7 Flight Planning Principles and image processing
Before beginning a flight project, the operator should ask and answer a se-
ries of sequential questions of drone design projects, approach, and preparations.
What is the size of the area to be mapped?
Depending on the equipment you have, larger the area to be mapped more
the choice goes to a fixed wing.
What is the size of the Ground Sample Distance (GSD)?
Small GSD values generally require lower flying heights and slower
ground speeds. For high-resolution projects (GSD 1 to 2cm) multi-rotors tend to
be the platform of choice.
What is the Nature of Terrain?
Multi-rotors are often referred to as Vertical Take-off and Landing
(VTOL), they are employed in congested areas such as forests and urban envi-
ronments.
What are the Flying Height and Ground Speed?
High ground speeds at low altitude require short exposure times to avoid
image blur. When GSD and focal length dictate a particularly low flying height,
the exposure distance intervals tend to be short, thus requiring the camera to ex-
pose at very short time intervals.
Ground sampling distance (GSD)
The spatial resolution of digital maps is commonly expressed as the
ground sampling distance (GSD). This is the dimension of a square on the
ground covered by one pixel (p) in the image and is a function of the resolution
of the camera sensor, the focal length (f) of the camera and the flying height (H
= the distance between camera and ground) (Barnes et al., 2014).
The formula is:
GSD H
p f
The pixel size (p) is found in the camera technical specifications, the di-
mensions of the image are specified in linear units (e.g. 17.3 x 13.0 mm) as well
as in a number of pixels (4000 x 3000 pixels). Pixel size is simply determined by
dividing the linear units by the number of pixels.
In photogrammetry to generate 3D models image capturing should be in
stereo modes which express an overlap between captured images. Once a flying
height has been determined it is necessary to compute the distance between each
exposure position, the spacing between flight lines, and the overlap (figure 7.1).

48
 Fig. 7.1: Image footprints and overlaps (Barnes et al., 2014).

The exposure distance intervals (s) and the spacing between the flight
lines (d) are dependent from the forward and lateral overlaps respectively. For
drone mapping projects we have found that a forward overlap of 80% and a lat-
eral overlap of 70% yield good results.
If the forward overlap is a% then:
s 100 (7.1)
Similarly, if the lateral overlap is b%, then:
100 (7.2)
Where β = GSD (width of the sensor in a number of pixels) and α = GSD
(length of the sensor in a number of pixels).
Some drone cameras cannot trigger at given distance intervals. In such
cases, the camera is programmed to expose at a fixed time. The time interval t
between two successive exposures is then calculated as
(7.3),
where d is the distance between exposures and v is the estimated ground speed
of the drone (Barnes et al., 2014).
Photocontrol points
Prior to commencing mapping from aerial photos, ground survey infor-
mation is required on specific terrain features in order to relate the photogram-
metric spatial model to its true geographical location. These terrain features may
be portrayed in two ways: by identifiable photo image features or ground targets.
Acquisition of ground control data on photo image points is a necessary
requirement for photogrammetric mapping for two primary reasons:
 To georeference the imagery.
 To check the accuracy of the spatial data collected.
Technology is constantly changing and adding to the tools that can be
used to collect ground control. Recent advances in GPS are the underlying rea-
son for many of the advances in ground survey methods, as well as photogram-
metry in general (Falkner, Morgan, 2002).
A ground target is some kind of a panel point that is placed on the unob-
structed ground prior to photography. Figure 7.2 illustrates the effective presen-

49
tation of a ground target (+) on an aerial photo. Ground targets create a discrete
image point and can perhaps lead to better map accuracy.

Fig. 7.2: Ground target on aerial photography.

A target must be of sufficient size to be recognized on the image. Dimen-

sions of a target in the shape of a cross are easily computed (Falkner & Morgan,
2002).
Equation 7.4 defines the width of the legs of a typical ground target, as
shown in figure 7.3.

0.002 (7.4),

where:
W = width of the target legs (cm),
Sp = scale denominator (cm).
Equation 7.5 defines the length of the target legs of a typical ground tar-
get, as shown in figure 7.3, for any photo scale.
10 (7.5)
Where:
l = length of each cross arm (cm),
w = width of target legs (cm).

50
 Fig 7.3: Typical ground target.

To produce the mapping from stereo models, the aerial photo image must
be scaled and leveled it means georeferenced to a true geographic ground loca-
tion. In current mapping procedures, most points contain both horizontal and
vertical information and are used for both scaling and leveling.
If the final mapping products are to be referenced to a defined coordinate
reference frame, then it is necessary to either geo-reference the aerial images or
set ground control points (GCPs). Ground control points are point features in the
object space that can be positively identified in the images figure 7.2. The tar-
gets are surveyed to determine their precise coordinates in a defined spatial ref-
erence frame. In most cases, ground control points are surveyed by means of
differential GNSS.
The Global Positioning System (GPS)
The global positioning system (GPS) an electronic receiver measures the
distances between the ground point and a minimum of four satellites and the in-
tersection of the divergent rays establishes the spatial coordinates of the observ-
ing station.
The anticipated ephemerides (positional) information is broadcast by the
satellite. Several continuous tracking stations are scattered throughout the world,
meticulously charting the paths of the satellites. Both the tracking data and
broadcast information are available to the user, with the former providing more
accurate data for processing receiver information (Falkner, Morgan, 2002).
The determination of the coordinates of a ground station by GPS proce-
dures relies upon intersection geometry. The GPS receiver, a pseudo-range
measuring device, accepts carrier signals from multiple satellites. By measuring
the carrier waves from a ground station to several satellite positions simultane-
ously, the XYZ coordinate can be determined. For ascertaining three-
dimensional coordinates, the receiver must maintain a continuous lock on a min-
imum of four orbiting platforms simultaneously (Falkner, Morgan, 2002).

51
 UAV’s usually use a standalone GPS receiver enabling the drone to navi-
gate safely and it is able to achieve a satisfactory accuracy for safe navigation,
the overall accuracy which can be achieved could not reach the survey grade
accuracy.
Static mode
In static traverses, at least two receivers must be used; both must be
locked on to the same group of satellites. One receiver resides over a location
with known coordinates, which could be a previous point in the circuit. The oth-
er receiver is set over a point with unknown coordinates. Observation time spent
at each baseline point pair may span a significant period of time. After the ob-
servation is complete at the baseline pair, both receivers can be moved to new
positions so long as one occupies a point with known coordinates. In this fash-
ion, observations are made at all of the stations on the traverse, in turn (Falkner,
Morgan, 2002).

Airborne Global Positioning

Today many photogrammetric mapping projects employ Airborne Global
Positioning technology to minimize ground control collection for the mapping
project. By locking on to several navigation satellites this device maintains a
constant spatial positioning record of the sensing systems. Relating the film co-
ordinates to a map allows the pilot to pinpoint the location of the aircraft on a
specific exposure frame. This coupling of remote sensing and spatial positioning
offers unique mapping capabilities.
Airborne Global Positioning technology has been developed to a point
where standard procedures and repeatable results make it almost standard prac-
tice for all medium to large project areas (Falkner, Morgan, 2002).
Many private and public organizations currently mount GPS receivers in
the airplane during a photo flight mission. This system allows the spatial fixing
of the camera at the instant of exposure.
Reference to Ground Station allows the Airborne Global Positioning re-
ceiver to be interfaced with a camera system. In this procedure, the GPS in the
aircraft is correlated with a static GPS ground station so as to relate the onboard
receiver with the known ground station.

Real-Time Kinematic mode (RTK)

kinematic traverses demand the inclusion of at least two receivers. One
receiver must remain fixed on a known ground station over the same point, dur-
ing the entire survey period. The other receiver onboard of the UAV giving cor-
rected coordinates of the exposure frame center.
By enabling a drone to become the ‘rover’ of a GNSS/GPS base station or
a Continuously Operating Reference Station (CORS) survey grade accuracy can
be achieved in real time at a sub-centimeter accuracy.

52
 RTK requires a constant communication link between the GNSS/GPS
base and the drone. In case this communication link is lost RTK will not work
anymore, to solve this problem the correction made by post-processing using
correction data coming either from a local base or CORS.
In RTK mode the receiver of the base station is dependent from the link
with the drone receiver trough UHF or mobile SIM card.

Post Processing Kinematic mode (PPK)

PPK surveys require data from at least two receivers: a reference base re-
ceiver and a moving rover receiver on board of the drone. If you have access
to raw data from a reference station (base or CORS).
PPK same as RTK requires a base station, in many countries cases the
public CORS
network is dense enough to provide corrections for the RTK and the PPK. It is
preferable to have your own base station. Using CORS comes with the possibil-
ity of connection loss, in both RTK and PPK, when the rover loses lock. The
advantage of PPK is that the search can proceed from previous and future data
relative to that instant.
There are two types of survey methods in PPK surveys: topo points and
continuous surveys. Topo points are short (about 15 seconds) occupations, e.g.
over a survey station. Continuous survey mode allows ongoing data collection at
a specified logging interval; continuous mode is used for drone mapping.
After drone landing, raw data collected from the mobile receiver (from the
drone) and the fixed receiver (from a base station), should be post-processed in a
computer software to produce an EXIF file containing images position data.
The accuracy requirement (horizontal and vertical) for the final mapping
datasets establishes the ground survey accuracy requirements and thus the GPS
methods.
The advantage of including techniques such as RTK (real-time kinemat-
ic) or PPK (post-processing kinematic) will:
- Eliminate the need for GCP’s, therefore reducing costs tremendously -
Reduce possible human errors.
- Achieve survey grade accuracy for the mapping/survey project.
- Makes it easy to proof accuracy by using a couple of checkpoints (Falk-
ner, Morgan, 2002).

Flight planning
In all drone projects before beginning the flight user should plan it first
following these planning steps:
1. Define the mission area, resolution and overlap.
Most professional drones employ waypoint navigation to define their
flight path. Modern flight planning software handles this automatically, incorpo-

53
rating the three pillars of aerial mapping: mission area, overlap and ground reso-
lution.
2. Generate a 3D flight plan.
When flying, the optimal flight planning approach is for the altitude of
each flight line.
An ideal flight planning system will have a high-quality elevation dataset
built-in, as well as allowing the operator's own elevation data to be imported.
3. Define take-off and landing points.
When defining the drone’s take-off trajectory, the software allows you to
define the exact direction the drone will follow after being launched. When
landing, choose the point and the position of landing, with its ground proximity
sensor, to adapt its altitude to the terrain, employing reverse thrust at an altitude
of five meters to reduce its speed and ensure a short, precise and soft landing.
4. Monitor and adapt during flight.
Even with intelligent, automated mapping drone systems, it is crucial to
carefully monitor every flight via line of sight and using the drone's ground sta-
tion software.
The drone's position and important parameters such as its battery's re-
maining charge, and therefore flight time, should always be visible on the screen
display (OSD) figure 7.4.

Fig. 7.4: Ground station on Screen Display of the telemetry, landing and post-flight tasks
(Barnes et al., 2014).

5. Launching the drone.

Drones are typically launched in one of the three ways: by hand, via roll-
ing take off on a runway, or using a launching system, most often a catapult.
6. During flight.
After take-off, a drone will quickly gain altitude and travel, either auto-
matically or on command, to take its mapping area and to start the aerial imag-
ing mission.
7. In-flight monitoring.

54
 Typically, an operator will ensure the safety of an operation by keeping
the drone in visual line of sight. The drone supports this human monitoring by
constantly sending status information displayed on the screen.
8. Landing.
After a mapping mission is finished, when landing, choose the point and
the position of landing, with its ground proximity sensor, to adapt its altitude to
the terrain, employing reverse thrust at an altitude of five meters to reduce its
speed and ensure a short, precise and soft landing.
It is good practice to inspect the UAV immediately after the landing for
any signs of overheating of the motors or electronics. The aerial images should
also be inspected for quality after the first flight and any required adjustments
made to the camera settings before further flights.

Flight planning software

Most flight planning software or computer tablet applications are auto-
mated and based on satellite imagery such as Google maps to help you locating
your mission and planning it. For flying heights below 100m, it is important to
inspect the terrain for objects, like power lines or cell towers, that are either not
visible or do not appear on the satellite imagery.
The first step for generation of a flight plan is drawing a polygon marking
the outline of the project area, putting the flight height, adjusting longitudinal
(forward) and lateral (side) overlap (%).
The flight planning software gives you the ability of flight plans modifica-
tion. One time drawing the flight plans in Google maps it automatically refer-
ences the flights to the WGS84 datum which is the same spatial reference frame
used by drones for navigation.
Flight planning software communicates with the autopilot on board of the
drone to receive broadcast positioning information from GPS satellite for posi-
tion updates and to send out control inputs to the motors of the drone. The most
common state observer is the inertial measurement unit (IMU) including gyros,
accelerometers, and magnetic sensors.

Image processing
The drone non-metric cameras are lighter and less costly they are general-
ly preferred over conventional metric cameras. However, non-metric cameras
are not geometrically stable, which is a basic requirement for classic photo-
grammetric mapping. To overcome this problem a combination of classic pho-
togrammetry and computer vision techniques, known as Motion Vision Stereop-
sis (MVS), is used by the most drones image processing software.

55
 Fig. 7.5: MVS workflow.

Normally drone processing software will attempt to find the approximate

camera geometry (focal length and sensor size). Note that in an MVS approach
the camera geometry for each and every image is computed separately, in this
way the camera exposure coordinates are computed in some arbitrary local spa-
tial reference system. Hence, if the final product does not have to be referenced
to some spatial reference frame, the images are all that is required to build a
model. However, the process of determining the relative camera exposure posi-
tions is significantly accelerated if the relative camera exposure positions are
seeded with the coordinates established in the after-flight image geo-referencing
step mentioned below.
Ground control is required to rectify (georeference) the imagery to its true
geographical position on the earth’s surface. The differential rectification meth-
od is a phased procedure which uses several XYZ ground control points to
georeference an aerial photograph to its true place on the earth. The exact loca-
tion and number of ground points required are based upon the scale and accura-
cy of the final orthophoto. Selecting the ground control points is generally not a
task for the project manager and should be decided by the photogrammetric
technician designing the project based upon his/her experience and equipment
(Falkner, Morgan, 2002).
The georeferencing is necessary since the photogrammetry software cre-
ates its own coordinate system when the models are generated. The selected
GCPs were marked in the processing software by placing markers in the photos
containing each point. The purpose of marking the points in multiple photos is to
allow the software to triangulate the locations of the point using the camera lo-
cations that have already been processed and gives a new georeferenced sparse
point cloud (Newsome, 2016).
Then the software uses the optimized sparse point cloud and improved
camera positions to generate a dense point cloud. Several accuracies and depth
filtering settings are provided in this process. The generation of the dense point
cloud usually takes the longest time of all the processing steps.
The dense point cloud is used to generate a surface model in the form of a
triangulated irregular network (TIN), a texture is built and applied to the surface
features of the model.
Various spatial products (Digital Surface Models and Orthophotos) can be
generated via the export options offered by the image processing software.

56
 Digital surface models can be prepared either in a vector format as trian-
gulated irregular networks (TIN), used extensively in CAD software for engi-
neering applications, or in a raster format for GIS uses.
A suitable DSM must be obtained to provide a vertical datum for an or-
thophoto, DSM for orthophoto rectification does not have to be as dense or as
detailed as a terrain model for contour generation (Falkner & Morgan, 2002).
Today, many uses for geospatial mapping products require current plani-
metric feature data. Analysis and design from geospatial data sets generally re-
quire a known positional accuracy of features. The collection and updating of
planimetric features in a data set can be costly. Many end users are also not ac-
customed to viewing and analyzing vector-based mapping data sets. They prefer
to view planimetric features like a photo image.
The final phase of the orthophoto process is the merger of the digital im-
age and the DSM along with corrections in pixel intensity throughout the image.
Software, used to merge the digital raster image with the DSM, makes adjust-
ments in the horizontal location of pixels based upon their proximity to DSM
points. This process removes the errors due to displacement and produces an
image that is orthogonally accurate (Falkner, Morgan, 2002).
Suitable imagery and ground control are the basic elemental data that de-
termines the final orthophoto reliability which involves both the accuracy of dis-
tances and areas within the orthophoto as well as the relative accuracy of fea-
tures with respect to their true location on the earth. Distance and area accuracy
is based on the pixel size. Relative feature precision is based on the accuracy of
the DSM used in the rectification process. The relative accuracy cannot be more
precise than the reliability of the DSM (Falkner, Morgan, 2002).
Drone mapping accuracy
The optimal condition for high accuracy mapping
 Work in soft lighting environment.
 Avoid transparent or reflective surface.
 Flight time around 11 am to 4 pm.
 Avoid Fisheye wide angle cameras because of the high distortions and
errors.
 Optimal coverage 80% overlap and 60% sidelap.
 Terrestrial, nadir and oblique images capture, must be processed in
different projects and then merge or unify in one project.
 Take GCP on transition images.
 Use manual stitch when needed.
 The use of drones with the highest precision GPS for not doing GCP.
 Use an image overlap of >85%.
 Set GCPs using a zigzag pattern.
 Add checkpoints to verify data quality.

57
  Batteries are affected by the cold good idea to pack spare, fully-
charged drone batteries.
 Remain aware from bird attacks.
 Avoid flying in high winds and thermal winds (uplift), rain and snow
(Reg Austin, 2010).
Ground Sampling Distance (GSD) accuracy
The quality of a project is linked to its accuracy, referring to the positional
accuracy not to be confused with camera resolution or a single image's Ground
Sampling Distance (GSD).
The ground sampling distance achieved depends on an imaging sensor's
pixel size and lens focal length, and is directly proportional to the drone's flight
altitude.
However, data output quality depends on much more than a camera's reso-
lution alone. (In fact, the dataset produced by a 12 MP camera will often be hard
to differentiate from that produced by a 16 MP model.)
DSM Accuracy Assessment
Accuracy assessment of the Digital Surface Model (DSM) is an important task
for horizontal and vertical assessments by comparing DSM with GCPs measured
by precise surveying instruments (GNSS receivers or Total Stations) in terms of
Root Square Error (RMSE). More specifically, assessments in Easting
(RMSEx), Northing (RMSEy) and vertical (RMSEz), horizontal (RMSExy), and
all components (RMSExyz) as suggested Aguera-Vera using equations as fol-
low:
RMSEx SQRT ∑ X X (7.6)
RMSEy SQRT ∑ Y Y (7.7)
RMSEz SQRT ∑ Z Z (7.8)
RMSExy SQRT ∑ X X Y Y (7.9)
RMSExy SQRT ∑ X X Y Y Z
Z (7.10)

Where XGCPi and XDSM are the X-coordinate component of GCP and cor-
responding coordinate in DSM, respectively; YGCPi and YDSM are the Y-
coordinate component of GCP and corresponding coordinate in DSM, respec-
tively; ZGCPi and ZDSM are the Z-coordinate component of GCP and correspond-
ing coordinate in DSM, respectively (Agüera-Vega et al., 2016).

58
 Chapter 8 Drones and Geospatial Analysis
There are a set procedures used for geospatial scientific research methods
divided in case studies and field work. Drone Field work is the collection and
gathering of information applied on the case studies of Multi-scale Digital Sur-
face Models, Multi-scale landforms classification, Multi-scale Terrain Rough-
ness, Terrain analysis for parcels evaluation, and Digital Surface Models to pla-
nar areas.
The geospatial analysis based on drone terrain datasets experiments car-
ried out at an altitude of 1700 meter above the sea level, in Zaarour region situ-
ated on the western Lebanese mountainous chain. We choose this bare non ur-
banized mountainous area because of its representative morphological terrain
forms.
The study area with a slight natural slope, represented by bare lands with
elements of anthropogenic relief. The inclusion of anthropogenic micro-relief in
the studying area due not only to the requirements of representativeness, but the
presence of complicating microform for the experimental modeling of the terrain
concave and convex smoothed areas.
An autopilot Dji Phantom 3 Unmanned aerial vehicle (UAV), caring a
camera of 14 megapixels at a focal length of 3.61 mm flown the study area at
different heights. The flight heights are measured from the takeoff point of the
quad copter; the experiment constituted from 6 missions: FA-20, FA-40, FA-60,
FA-120, FA-240 and FA-360 of 20, 40, 60,120,240 and 360 meters height.
The flight path followed by the quad copter was identical for all the flights
and it was designed in a mobile application called Litchi. The application shows
the flight path and the flight parameters (coordinates, height, time, etc…).
Before starting the aerial surveying, well-distinguishable 10 control points
were evenly distributed within the area of interest for scaling and georeferencing
the resulted data. Ground control points (GCP) were collected with Global Posi-
tioning System (GPS) in stereographic coordinate system.
The drone took Aerial photography with 80% overlapping and 70% side
lapping. SfM-based 3D methods operate on the overlapping images. The drone
flight in an autonomous way, defined by waypoints to avoid image coverage
gaps, every surface that will be reconstructed needs at the minimum to be cov-
ered by at least 2 images taken from different positions.
All datasets (photos) of the six missions of different flight heights was
processed in Agisoft photoscan software for the generations of Digital Surface
Models (DSM) and Digital Ortho Model (DOM), when the camera focal length
and the flying height of the UAV are known the scale is determined by this for-
mula:
1

With S =scale, f = focal length of the camera, H= flying height.

59
 The result of scale calculations for each flight Height is listed in table 8.1.

Table 8.1: The spatial resolution, calculated scale and approximated scale of each DSM type.

DSM Spatial resolu- Calculated Scale Approximated Category

tion (m) Scale
FA-20 0.37 1/5540 1/5000 Plans
FA-40 0.55 1/11080 1/10000
FA-60 0.80 1/16620 1/15000
FA-120 1.73 1/33241 1/30000 Maps
FA-240 3.20 1/66481 1/60000
FA-360 4.47 1/99722 1/100000

Figure 8.1 shows six DSM of the study area of different spatial resolu-
tions: FA-20 of 20 meters flight Height with a very high resolution data set high-
lighting all the terrain details even rock textures, passing by FA-60 and ending
by FA-360 of 360 meters’ flight Height with a very low spatial resolution and
high generalization effect with the «disappear» and «growth» of some terrain
morphological features.

Fig. 8.1: Multi-scale DSM extracted from UAV photogrammetry methods.

60
 These 6 DSM can be classified visually from figure 8.1 by rough and
smooth; FA-20, FA-40 and FA-60 for rough and FA-120, FA-240 and FA-360
for smooth, also figure 8.1 constitute an interval of scales and smoothness show-
ing the generalization at different scales.
As per table 8.1 different flight Height lead to different spatial resolution
(pixel size); the minimum spatial resolution is 0.37 m which is a high resolution
showing all terrain details and a maximum resolution of 4.47 m quite good reso-
lution for geomorphological analysis at a local scale.
After the scales calculation we categorized our data in two categories
plans and maps; the first three DSM of flight Height 20, 40 and 60 are related to
the category of plans and the other are related to the category of geographical
maps.

Multi-scale Digital Surface Models

In geoinformatics, scale is predominantly considered a function of the
resolution of Digital Elevation Models (DEMs) (Hengl and Evans, 2009; Mac
Millan and Shary, 2009).
The fast progress in technologies especially the unmanned aerial vehicles
(UAV) and new photogrammetry softwares encourages acquisition and pro-
cessing of Digital Surface Models (DSM) at ever finer resolutions. The scale
dependency of land-surface parameters was noted by Evans (1972) as «a basic
problem in geomorphometry» (Shary et al., 2002). Meanwhile, the scale de-
pendency of land-surface parameters and land-surface objects has been
confirmed by several researches (Chang, Tsai, 1991; Wood, 1996; Florinsky,
Kuryakova, 2000; Evans 2003; Hengl, 2006; Arrell et al. 2007; Deng et al.,
2007; Pogorelov, Doumit 2009; Wood, 2009).
The factor of scale and resolution plays an important role in the uses of
digital models in GIS research, the scale may range from micro to macro scale
and may be in different levels of measurement, this study will only discuss the
spatial scale with the UAV flight Heights. For the selection of an image with
appropriate spatial resolution for a study demand the characteristics examina-
tion, especially the changing pattern as a function of changes in scale and resolu-
tion.
As per Li (1993), based on different kinds of philosophy, three types of
approach can be identified to the generation of multi-scale representations for a
given DTM (Digital Terrain Model), i.e. critical-points-based, smoothing-based,
and scale-driven (Li, 1993).
Woodcock and Strahler (1987) suggested the use of the local variance
method to find images with the optimum scale and resolution.
In their experiment, Woodcock and Strahler (1987) used image data were
degraded to coarser spatial resolution by resampling the neighbors’ cells and the
pixel value of the coarse resolution is an average of a group of finer resolution.
In our experiment we have 6 DSM for the same study area with different spatial

61
resolution obtained from UAV different heights survey. The effect of scale (spa-
tial resolution) on these surface models is analyzed according to the morphomet-
ric indices by calculating their direct indicators of spatial correlation. Therefore,
the experiment implemented the opportunities of multi-scale measurement tech-
nology based on UAV.
For multi-scale analysis we applied Wood and Strahler (1987) method of
Local Variance, the texture method measuring surface properties such as
coarseness and smoothness, and as a third index of fractal dimension.
Many scientists, quantitively discussed the issue of optimal resolution of
digital elevation models. Our study summarized the methods expressing relation
between UAV flight Height and spatial resolution, and they are local variance,
texture analysis and the fractal method. The first two methods are relative sim-
ple and very useful in practical sense, while the third fractal dimension method
has a great potential in detecting the resolution effects and is used especially in
several geosciences researches.
Local variance method proposed by Woodcock and Strahler (Woodcock
and Strahler 1987) is such a method, originally developed in image analysis,
with potential for dealing with scale in DEM analysis (Li, 2008). Local variance
measures the mean of standard deviation within a 3 by 3 pixels in a moving
windows, the mean of all local standard deviation values over the entire image
were then used as an indicator of local variability contained by the image.
The local variance is the degree of similarity between the values of two
points depending on the spatial distance between them. The longer the distance
is, the higher the degree of similarity is, and the smaller the variance is.
According to Schmidt and Andrew (2005), the land surface is hierarchi-
cally structured and it can be represented differently across scales for example
convex hillslope embedded into a concave hillslope, which in its turn is embed-
ded into a valley. These cases could be detected and seen only on high spatial
resolution DSM and are homogeneous relative to scale levels.
In figure 8.2 the variance maps of the study area at FA-20 showing all
morphological forms in detail even small concave and convex forms with some
finesse in FA-40, the same forms became bigger in size and dimensions. In the
last stage of plan categories FA-60 the ridges boarding the roads became more
highlighted with disappearance of the small morphological forms. Hence in the
second category of scales FA-120, FA-240 and FA-360 the degree of smooth-
ness is increasing with the scale.

62
 Fig. 8.2: Variance digital models of the six flight Heights.

Texture is a measure of coarseness, smoothness and regularity of the sur-

face (Gonzalez, Wintz 1987); texture analysis can measure the spatial variability
of image data, and can improve the statistical separation of the otherwise simi-
larity reflecting surfaces (Nellis, Briggs, 1989). The differences in texture for
images covering the same area with different scales and resolutions (the case of
our study) can indicate the heterogeneity of the scene under observation. In that
way we can compare the data between each other, in this case highest texture
index indicates the highest variation.
Texture is calculated by extracting grid cells that outline the distribution
of valleys and ridges. Each grid cell value represents the relative frequency (in
percent) of the number of pits and peaks within a radius of ten cells (Iwahashi,
Pike, 2007).
Similar to local variance methods the texture analysis method is also
based on the variability of the geographic data change of scale and resolution,
and the scale at which the maximum variability occurs is where most of the re-
lief formation processes operate. By finding the maximum variability of the da-
taset, we could find the operational scale of the geographic phenomenon and
therefore we could decide the observational scale of the study.
The texture index in figure 8.3 shows a stability in the morphological
forms in the first three scale and an unclear form in the last three level of scales.
Consequently, the first three scales allow us to determine the Genesis of the re-
lief-forming processes.

63
 Fig. 8.3: Texture digital models of the six flight Heights.

Fractal dimension, the concept of fractals fascinates geographers because

all the patterns of nature are too irregular and too fragmented to be quantified
using the traditional measure of geometric shapes. In real world curves and sur-
faces are pure fractal that’s why fractal could play an important role in detecting
the scale and resolution effects in remote sensing and GIS.
Eastman (1985) developed a single pass technique for measuring the frac-
tional dimension of lines. The procedure considers each slope segment, to pro-
vide evidence of an underlying angularity that can be considered as the generat-
ing angle of the fractal form. The formula is based on calculated slopes as fol-
lows:
log(2)
D (8.1)
180  slope
log(2)  log(sin( ))
2
We applied Eastman’s method to calculate the fractal dimension of the six
DSMs (Figure 8.4). As it is known, surface fractal dimension values vary inside
the interval between 2 and 3. Empirical studies indicate that the fractal dimen-
sions of curves and surfaces change with the scale and resolution, the scale at
which the highest fractal dimension is measured may be the scale at which most
of the processes operate (Goodchild, Mark, 1987; Lam, Quattrochi, 1992).

64
 Fig. 8.4: Fractal dimension digital models of the six flight Heights.

The fractal dimensions show the roughness and complexity. Goodchild

(1980) found that fractal dimension could be used to predict the effects of carto-
graphic generalization; many scientists used fractals to characterize landscape
environmental data, soil, topography (Burrough 1983; Mark, Aronson 1984; Po-
gorelov, Doumit 2009).
The fractal dimension of an image is expected to be lower as the resolu-
tion becomes coarser; because coarser resolution is likely to result in low varia-
bility in digital surface model, it is argued that the best resolution level for a
study is the one that has the highest fractal dimension within a stable scale
range, the results of fractal dimensions calculations shown on figure 8.4.
If we compare figure 8.2 the local variance index maps with figure 8.4 of
fractal dimensions a high degree of similarity is seen, which indicates the use-
fulness of the application of indices in this analysis.
After the application of the three scale indices on the six levels of details a
statistical table summarizing in values the generalization between them.

65
Table 8.2: Local variance, texture and fractal dimension statistical values over the six DSM’s.

Local Variance Texture Fractal Dimension

DSM
Min* Max Mean STD Min Max Mean STD Min Max Mean STD
FA-20 0 0.129 0.005 0.007 0.189 0.864 0.457 0.170 2.0 2.179 2.008 0.011
FA-40 0 0.136 0.010 0.012 0.000 0.851 0.337 0.182 2.0 2.112 2.008 0.009
FA-60 0 0.143 0.015 0.017 0.000 0.75 0.333 0.157 2.0 2.067 2.006 0.007
FA-120 0 0.458 0.088 0.065 0.009 0.561 0.221 0.125 2.0 2.042 2.007 0.006
FA-240 0 0.811 0.222 0.141 0.160 0.408 0.252 0.059 2.0 2.023 2.006 0.004
FA-360 0 0.985 0.414 0.231 0.211 0.401 0.295 0.050 2.0 2.016 2.005 0.003
*Min – minimal value, Max – maximum value, Mean – mean of the values, STD – standard
deviation of the values.

The local variance maximum values increase with the flight Height and
the scale, the standard deviation and the mean are increasing proportionally in
all scales, contrary to local variance values, texture and fractal dimensions max-
imum, mean and standard deviation are decreasing with the flight Height and the
scale as a result of the smoothing effect.
To determine the spatial distribution of the analyzed indices at different
scales, we applied a simple and effective tool – spatial correlation.

Fig. 8.5: Correlation scatterplots of the DSMs values.

A correlation scatterplot of all the elevation values of each DSM com-

pared to the main DSM the FA-20, from figure 8.5 we can see a good correla-
tion between FA-20 and FA-40 (coefficient of determination R2 is equal to
78.7%), which decreases between DSM FA-20 and FA-60 (R2=65.4%).

66
 The scatterplots of the last three levels show a very poor degree of simi-
larity comparing to FA-20 in fact, it shows a lack of connection between DSM
at various levels of scale. It means we are losing accuracy and we have a big
generalization lost in the maps category, built using the UAV. Similar results
were obtained when comparing the estimated parameters of the texture.

Table 8.3: Correlation analysis values between different scales of different indices.

Values of R2%
DSMs DSM Fractal Local Variance Texture
FA-20/FA-40 78.71 63.9 71.79 78.68
FA-20/FA-60 65.39 24.95 25.09 65.28
FA-20/FA-120 40.03 27.04 26.07 39.49
FA-20/FA-240 14.01 20.72 18.39 13.15
FA-20/FA-360 0.78 13.55 11.89 0.79

When interpreting the values of local variance, we consider the following:

If the spatial resolution is considerably finer than the objects in the scene, most
of the measurements in the image will be higher correlated with their neighbors
and the local variance will be low. If the objects to be studied approximate the
size of the resolution cells, the value tend to be different from each other and
therefore the local variance increases. The resulted correlation values in table
8.3 of DSM and texture are decreasing proportionally with the scale otherwise
FA-20/FA-120 at fractal and local variance constitute a barrier of values due to a
raised amplitude in R2 values which give a similarity in scale changes between
fractal and local variance in a way and in another way between DSM and Tex-
ture.
The correlation between the values of the local variance has a statistical
reliability (significance) only for the DSM with a spatial resolution close to FA-
20 and FA-40. For cases with large differences in spatial resolution, these indi-
ces become essentially independent on the same site.
We can conclude that the correlation in local variance decrease with flight
Height.
Scale and resolutions effects have been and will continue to be an im-
portant issue in geographic research. It is essential to have a good understanding
of the effects of scale on the analysis results. We tested in this study three mor-
phological indexes (local variance, texture, fractal dimension) that can be used
to detect the scale and resolution effects.
The results showed a weak relationship of spatial characteristics to com-
pare the images to scale plans and maps. Simultaneously, the survey showed that
the characteristics of the local variance and texture are good indicators to deter-
mine the optimal spatial resolution of the morphometric analysis of the Earth's
surface.

67
 Multi-scale landforms classification
The factor of scale plays a very important role in Landform classification
different levels of measurement (nominal, ordinal, interval and ratio) this study
will discuss the terrain analysis with the applications of Terrain Position Index
(TPI), Iwahashi and Pike index and the morphometric features and their effects
on generalization and spatial resolution at different UAV flights altitudes.
Pike et al. (2009) remarked that no digital elevation models derived map
is definitive, as the generated parameters differs with algorithms and can vary
with resolution and scale.
Landform classification stand out with terrain complexity which necessi-
tated specific methods to quantify its shape and subdivide it into more managea-
ble components (Evans 1990) which constitutes a central research topic in geo-
morphometry (Pike 2002; Rasemann et al., 2004).
An Arc Map Jenness module GIS software for landforms terrain computa-
tions was applied on a three different spatial resolutions drone based DSM’s for
the extractions of Topographic Position Index (TPI), Iwahashi and Pike land-
forms and the morphometric features at different scales.
Throughout the assessment, we comprehensively used this UAV for aerial
images acquisition to the generation and interpretation of Digital Surface Mod-
els (DSM) by using new photogrammetry technologies.
A three DSM of different spatial resolutions, FA-20 of 20 meters’ flight
altitude with a high resolution highlighted all the terrain details even rocks tex-
ture, passing by FA-120 the terrain is smoothed with some concave and convex
areas and ending by FA-360 a very low spatial resolution and a very smoothed
terrain of 360 meters’ flight altitude.
Topographic Position Index (TPI), the analysis was performed by
DSM’s simulation to obtain topographic position index (TPI). The process of
formulae (1) calculate the difference between elevation at a specific cell and the
average elevation of the neighborhood surrounding cells (Tagil & Jenness
2008); describing higher and lower areas for the classification of the terrain into
different morphological forms (Jenness 2005).
The simulation required the radius adjustment of neighborhood and its
geometric shape based on two different scales or two sizes (Barka et al. 2011).
In this study, a radius between 5 m and 25 m was applied to determine the slope
positions.
∑
TPI Z (8.2)
Where:
Z0 = elevation of the model point under evaluation,
Zn = elevation of grid within the local window,
n = the total number of surrounding points employed in the evaluation.
These neighborhood radiuses values were applied for all DSMs spatial
resolutions, to be similar in parameters for best comparison analysis.

68
 Positive TPI values represent high locations e.g. ridges, negative values of
TPI represent low terrain representations e.g. valleys otherwise flat areas have
TPI values near zero, high positive values go to high elevations geomorphologi-
cal structures such as peaks and ridges (Jenness, 2010).
The flight altitude FA-20 has a maximum positive value of 1.03, FA-60 of
0.71 and the higher flight altitude FA-360 with 0.48 a decreasing in maximum
and minimum values with the increasing of flight altitude.
Iwahashi and Pike had developed a Landforms classification unsuper-
vised method based on only three terrain attributes: slope gradient, surface tex-
ture and local convexity (Iwahashi and Pike 2007). This method restricts a num-
ber of landform classes 8, 12 or 16 with a physical meaning of statistical land-
scape properties.
The unsupervised approach treats topography as a continuous random sur-
face, especially for the three level of details FA-120, FA- 240 and FA-360 inde-
pendent of any spatial or morphological orderliness imposed by fluvial activity
and other geomorphic processes.
Morphometric elements, the standard method for the identifcation mor-
phological elements is to establish a mutually position for the central cell in rela-
tion to its neighbors (Peucker, Douglas 1974; Evans 1979). The classification
algorithm can be done by maintaining the continuity of linear elements, which
gives advantages over the method of selection on the basis of logical comparison
of neighboring cells (Peucker, Douglas, 1974; Jenson, Domingue 1988; Pogore-
lov, Doumit, 2009).
Morphological elements take the forms of: Planar, pit, channel (thalweg),
pass, ridge (division line), and peak. The names of morphological elements may
vary in different sources, but they can be uniquely explaining in terms of chang-
es in the three orthogonal components x, y and z (Wood J, 1996; Pogorelov,
Doumit 2009).
Landform classifications delineated using the TPI method is shown in fig-
ure 8.6, TPI values present a powerful way to classify the landscape into mor-
phological classes (Jenness, 2005). Landform Classifications consist of “Can-
yons, Deeply Incised Streams’’, ‘‘Midslope Drainages, Shallow Valleys’’ and
‘‘Upland Drainages, Headwaters’’ all tended to have strongly negative curvature
values of a concave shape, while “Local Ridges or Hills”, ‘‘Midslope Ridges,
Small Hills in Plains’’ and ‘‘Mountain Tops, High Ridges’’ all tended to have
strongly positive curvature values of a convex shape.
Figure 8.6 of the three maps shows land forms classification of all mor-
phological forms listed above at different scale level, a visual analysis of these
maps highlight a cartographic generalization between them making a very clear
evolution in morphological forms at each stage.

69
Fig. 8.6: maps of landform elements of the three DSM derived from TPI classification analy-
sis. a) FA-120, b) FA-240, c) FA-360.

The results of table 8.4 shows how the area of some morphological ele-
ments is increasing against other elements relating to scale variations. In table
8.4 the area percentages of some morphological elements are increasing in val-
ues and other are decreasing with the scale variation. streams, plains, open
slopes and high ridges are increasing in area and geometrical forms due to the
variations in spatial resolution. Some morphological elements such as Upland
drainage type are not found in any of the three maps and other like Local ridges
are disappearing with scales variation and constituting a basic for generalization
processes.
Open slopes comprised between 6 and 11% of the total area in all flight
altitudes while midslope drainages increasing with the flight heights between
9.75% and 11.46% from the total study area.
Landforms show a decreasing in their numbers; the dilution of 647886
pixels of different morphological elements from the flight height FA-120 to the
flight height FA-360. All the ten morphological elements are affected by scale
generalization.
Open slopes comprised between 6 and 11% of the total area in all flight
altitudes while midslope drainages increasing with the flight heights between
9.75% and 11.46% from the total study area.

70
Table 8.4: Percentage of Morphological elements and pixels numbers of each Morphological
element in the three DSMs levels based on TPI classification.

Area (%) Number of pixels

Type FA-120 FA-240 FA-360 FA-120 FA-240 FA-360
Canyons, deeply incised 5.87 9.32 14.7 6155 1036 800
streams
Midslope drainages, shal- 9.75 10.6 11.46 31823 6684 5984
low valleys
Upland drainages, head- 12.82 11.58 9.7 79684 19287 14025
waters
U-shaped valleys 17.24 13.02 9.01 112465 29124 18901
Plains 13.1 9.69 6.39 158603 42986 22098
Open slopes 10.94 9.1 6.35 153612 42504 21892
Upper slopes, mesas 10.9 10.77 8.87 110730 30022 18910
Local ridges, hills in val- 7.99 9.49 9.53 76494 19574 14150
leys
Midslope ridges, small 6.4 8.88 10.72 35575 6902 6227
hills in plains
Mountain tops, high 4.99 7.55 13.28 6818 1663 1086
ridges
Total numbers of pixels 771959 199782 124073

Landforms show a decreasing in their numbers; the dilution of 647886

pixels of different morphological elements from the flight height FA-120 to the
flight height FA-360. All the ten morphological elements are affected by scale
generalization.
To understand the degree of generalization between the big scale of FA-
120 and the small scale of FA-360 we provided an ascending classification of
the geomorphological forms, ridges (Local, Midslope and high) then drainage
areas (upland and midslope), hence all other morphological forms are positively
affected by generalization by raising their areas.

71
Fig.8.7: landform maps of unsupervised classification (Iwahashi and Pike method),
a) FA-120, b) FA-240, c) FA-360.

Table 8.5: Iwahashi and Pike landform percentage of areas at different scales.

Area (%)
Type FA-120 FA-240 FA-360
1) very steep slope, fine texture, high convexity 0.00987 ─ ─
2) very steep slope, coarse texture, high convexity 47.33649 48.5 49.9
3) very steep slope, fine texture, low convexity 0.00144 ─ ─
4) very steep slope, coarse texture, low convexity 51.13720 51.2 49.9
5) steep slope, fine texture, high convexity ─ ─ ─
6) steep slope, coarse texture, high convexity ─ ─ ─
7) steep slope, fine texture, low convexity ─ ─ ─
8) steep slope, coarse texture, low convexity ─ ─ ─
9) moderate slope, fine texture, high convexity ─ ─ ─
10) moderate slope, coarse texture, high convexity ─ ─ ─
11) moderate slope, fine texture, low convexity ─ ─ ─
12) moderate slope, coarse texture, low convexity ─ ─ ─
13) gentle slope, fine texture, high convexity ─ ─ ─
14) gentle slope, coarse texture, high convexity 0.67723 0.2 0.1
15) gentle slope, fine texture, low convexity ─ ─ ─
16) gentle slope, coarse texture, low convexity 0.83771 0.2 0.1

72
 The concavity and convexity of the very steep slope with fine texture
found only in high spatial resolution models (FA-120), the coarse texture of high
convexity increasing with the pixel size.
Step and moderate slopes are not detected in all three models, gentle slope
coarse texture high and low convexity are increasing with the flight altitude.
Varying DSM spatial resolution can achieve an elements separation of
appropriate scale, without the need of generalization.

Fig. 8.8: Morphometric features maps, a) FA-120, b) FA-240, c) FA-360.

Table 8.6: Surface specific points area percentages of the study area at different scales.

Area (%)
Type FA-120 FA-240 FA-360
Planar 0.00001 ─ ─
Pit ─ ─ ─
Channel 49.72501 48.3 47.4
Pass (saddle) ─ ─ ─
Ridge 50.27497 51.7 52.6
Peak ─ ─ ─

As per table 8.6 some morphometric features like pit, pass and peak are
not detected in all flight altitudes, otherwise planar areas are detected in FA-120
the lower flight altitude at a very low percentage of area in order of 0.00001 %,
we cannot judge on this result because the value of this pixel could be a pro-
cessing artifact. The area of channels is increasing with the flight altitude and
the ridge area is decreasing against the channel one.

73
 The dominating land forms of surface specific points channel and ridges
of the study area form a comparison models of each flight height with TPI land
forms. By splitting channels and Ridges of FA-120, FA-240 and FA-360 and
exanimating which TPI land forms are included in each type, table 8.7 shows the
area percentage of each landform.

Table 8.7: Percentage of TPI landforms containing in Ridges and Channels

at each flight altitude.

Percentage of area
TPI Landforms Ridge- Channel- Ridge- Channel- Ridge- Channel-
120 120 240 240 360 360
Canyons, deeply in- 1.5 10.4 2.4 16.8 4.3 26.3
cised streams
Midslope drainages, 3.3 16.3 3.9 17.8 4.9 18.6
shallow valleys
Upland drainages, 5.7 19.7 5.4 18.0 5.6 14.4
headwaters
U-shaped valleys 11.9 23.1 9.3 17.1 7.0 11.2
Plains 13.7 12.4 9.6 9.6 6.3 6.5
Open slopes 14.7 7.1 11.4 6.8 7.1 5.4
Upper slopes, mesas 16.5 5.3 15.3 5.7 11.5 6.0
Local ridges, hills in 12.9 3.0 14.9 3.7 13.8 4.9
valleys
Midslope ridges, small 10.7 1.9 14.6 2.7 16.8 3.9
hills in plains
Mountain tops, high 9.1 0.9 13.1 1.6 22.7 2.9
ridges

Upper slopes areas in ridge-120 and ridge-240 occupied a high percentage

of areas, for ridge-360 the higher percentage of area goes to Mountain tops. Up-
land drainage owns high values in channel FA-120 and FA-240 but for FA-360
the higher area goes to Canyons. From these results we can see a TPI landform
transition with scales.

74
 Ridge
25

20

15

10

Ridge-120 Ridge-240
Ridge-360 Логарифмическая (Ridge-120)
Логарифмическая (Ridge-240) Логарифмическая (Ridge-360)

Fig. 8.8: diagram of area percentage of TPI landforms containing in Ridge

at flight altitudes of 120,240 and 360 meter.

The diagram of figure 8.8 shows the percentage of TPI land forms area in
ridges at different scales, the log curves of 120, 240 and 360 have an intersection
point at upper slope this point made a transition of values from low percentage
to higher percentage of areas.
The correlation value of R2 between land forms of FA-120 is 0.6 for FA-
240 is 0.9 well correlated because of the proportionality and small percentage
interval of areas, for FA-360 the correlation value is 0.6 similar to FA-120.

75
 Channel
30
25
20
15
10
5
0

Channel-120
Channel-240
Channel-360
Логарифмическая (Channel-120)
Логарифмическая (Channel-240)
Логарифмическая (Channel-360)

Fig. 8.9: Diagram of area percentage of TPI landforms containing in Channel at flight alti-
tudes of 120, 240 and 360 meter.

Channel usually are concave areas, in figure 8.9 we can see dominating
the area of canyons in FA-360, the correlation of area percentage between the
landforms of FA-360 is very high with R2= 0.97 and a concave logarithmic trend
line.
Otherwise for FA-240 a less concavity logarithmic trend line with R2=
0.75 due to the proportional percentage of areas between landforms.
Fa-120 has a low correlation between landforms R2= 0.35 even less than
the average.
We can conclude from these values that due to cartographic generalization
and the transition from flight altitude to other, the degree of similarity for chan-
nels landforms areas rising with the flight altitude. Hence for ridges land types
the area of canyons and midslope, upper slope local ridge, midslope ridge and
mountain tops are increasing with flight altitude, upland drainage, u-shaped val-
ley, plain and open slope area is decreasing with the flight heights.
By using Topographic Position Index and unsupervised classification of
Iwahashi and Pike, the study area was classified into landform categories of dif-
ferent scale DSM. The result shows that ridges and drainage forms are more af-
fected to generalization than other forms.

76
 The landform classes obtained for the three scales differentiate dynamic
terrain characteristics of the study area. Landform classifications extracted form
drone DSM and GIS fast the presented results and discussion by integrating the
geospatial multiscale approach of terrain analysis.
The result shows that TPI provided a powerful tool for describing topo-
graphic attributes of a study area and there is a relationship between landform
map and spatial resolution. By deep understanding of the terrain characteristics,
potential and specific constraints of cartographic generalization. Information and
methods discussed in this study are valuable results for cartographic multiscale
studies and analysis. Landforms are dissolving with scales against each other’s,
some of them gaining areas and some disappeared. This study analyzed the gen-
eralization at three different scales (flight altitude), for future researches we are
planning to examine and monitor changes of landforms at micro, local and glob-
al scales.

Terrain analysis for parcels evaluation

Terrain criteria means geomorphometrical parameters or land terrain
quantitative analysis, elevations interval, slope and terrain curvature are the key
for parcel evaluation.
Topographic and cadastral maps give an idea about the boundary and the
terrain structure, for best reputation some experts took into account terrain forms
in parcels evaluation.
With the appearance of geoinformation technologies applied to real estate,
terrain analysis became very easy. Unmanned aerial vehicles (UAV) combined
with Computer Vision and Geomatics knowledge, allowed the generation of
horizontal and vertical terrain spatial information with positional quality at the
centimeter level. Structure from Motion (SFM) a modern tool for UAV's images
processing (Westoby et al., 2012) has been used for data generation. This ap-
proach, implemented in many processing photogrammetry softwares (Pho-
toScan, VisualSfM, and others). Horizontal information expressed by Digital
Ortho Model (DOM) or simply aerial images and the vertical terrain spatial in-
formation is a Digital Surface Model (DSM).
In this study at an experimental land in Zaarour region (Lebanon) an ele-
vation, slope and curvature, derived from the DSM processed and integrated in a
GIS spatial terrain weighted analysis to output a parcel evaluation map.
The mean values of the resulted evaluation map within the parcels bound-
aries are used for ranking and real estate arrangement.
Datasets were collected at a flight height of 360 m above the ground from
the taking off point, resulting a swath of 556×456. The images forward and side
overlaps were approximately 90% and 70% respectively.
The final products of Photoscan standard workflow, described by manu-
facturer (Agisoft, 2013) are a points cloud, 3D model, an orthophoto and a digi-
tal surface model.

77
Fig. 8.10: A) terrain representation DSM of the study area, b) true color Digital Ortho Model.

The DSM generated from Photoscan in figure 8.10 a shows the terrain
structure (ridges and channels) and elevations varying from 1689 to 1821 meters
above the sea level. Terrain texture and pattern in figure 8.10 b highlight a very
smoothed bare land and some manmade traces as roads and ponds.
Usually real estate experts evaluate lands by their terrain, nowadays GIS
modules and algorithms are a very important tool for evaluation and decision
making.
Terrain analysis and land assessment based on DSM resolved: local relief,
slope and terrain curvature for parcels real estate evaluation.
The local relief is defined as the difference between the highest and low-
est elevations occurring within that area. Local relief was introduced by Partsh
(1911), Evans (1972) compared the values of local relief determined over more
than one size of area, and he recommended the use of a fairly large sample area.
In our study we calculated the local relief inside each parcel with a calcu-
lation of all elevation statistic values (minimum, mean, and maximum).
The local relief of a big parcel shouldn’t act same as local relief of a small
parcel that’s why we divided the local relief value by the area in square meter
and multiply it by 100 to reduce decimals value equation 3.

100 (8.3)
Where:
Emax – Maximum elevation,
Emin – Minimum elevation,
A – Parcel area.

78
 The values of V express the difference in elevations related to parcels ar-
ea, they are in the range of 1 to 59, high values went to parcels with low local
relief and big areas for high local relief values.
The Slope is one of the most fundamental assessments of parcels and
landscape characteristics, and is reported as a driving variable in many construc-
tion studies. A parcel with an extreme slope is bad evaluated in real estate mar-
ket, very difficult accessibility and not useful for investment.
The mountainous location of the study area and the nearest to Zaarour
country club for winter games gives an idea about the high slopes of the region.
A variation of slope from 0.8 % plain terrain to 96 % rocky cliff structures char-
acterized the evaluated parcels. The variety in parcels slopes gave a variety in
real estate prices.
Terrain curvatures computation is very complicated because, in general,
the surface has different curvatures in different directions. With GIS technology
and software, terrain curvature became a very easy parameter to calculate, the
parameter describing the concavity and convexity of the surface is called curva-
ture, which give a proper assumption about the nature of the land surface with in
the parcel boundary.
Based on Digital Elevation Models, the most popular algorithms for deriv-
ing terrain curvature are those of Evans (1972), Shary (1995), Zevenbergen and
Thorne (1987), as well as the modified Evans-Young (Shary et al., 2002) meth-
od Burrough and McDonnell (1998) gave preference to the Zevenbergen-Thorne
algorithm. For the extraction of the parcels terrain curvatures, we used the
Zevenbergen-Thorne’s method which is based on the idea of representation of a
surface by an equation of partial quadratic form (Zevenbergen, Thorne, 1987).
The Zevenbergen and Thorne module applied to the DSM generated from
the photogrammetric processing shows that the concave surface of the study ar-
ea is 14%, 53% convex and 33% flat areas. These results can be the prove that
must of the parcels are hilly as per the DSM 8.10 a.
All listed above DSM extracted maps constitute criteria for real estate
land evaluation map, McHarg describes the process of conducting multi-criteria
analyses by categorizing and ranking values from a variety of thematic datasets,
creating a transparency for each dataset, and then overlaying the transparencies
together to create a composite image. This final composite image was then used
to evaluate the suitable land uses in the design scenario (McHarg, 1995).
The value of this weighted overlay approach was understood and adopted
by many researchers, who saw great potential from the early GIS programs.
The weighted overlay approach allows researchers to create composite
maps, in our project a parcel evaluation map for real-estate assessments. These
composite maps made it possible for decision makers to consider multiple at-
tributes in a single map (Hopkins, 1977).

79
 Multiple Criteria Decision Making (MCDM) methods designed to help
stakeholders (real estate experts) make well-informed decisions based on vari-
ous attributes (Jankowski, 1995).
Each criterion evaluation of local relief, slope and curvature takes a value
from: 1 to 5 number of classes, 1 being the worst value and class 1 the worst
class, otherwise the class number 5 is acting as the best value.

Table 8.8: Weights of Local relief, slope and curvature classes.

Local relief Slope Curvature

Classes Score Classes Score Classes Score
0.04-0.15 5 0-14 5 -110.37--12.07 1
0.15-0.17 4 15-29 4 -12.07--3.52 2
0.17-0.27 3 30-44 3 -3.52-2.1 5
0.27-0.70 2 45-59 2 2.17-12.14 4
0.70-2.52 1 >60 1 12.14-71.26 3

Local relief and curvature raster datasets were classified using the geo-
metric interval classification of ArcMap method. This classification method was
used for visualizing continuous data, the specific benefit of the geometrical in-
tervals classification is that it works reasonably well on data that are not distrib-
uted normally.
Local relief calculated by formulae 1 gives a range of values from 0.04 to
2.52, the small values with small interval of elevation and high values with big
interval the highest score went to the small values and gradually going down to
1 for the interval of big values.
Curvature negative values are concave terrain forms and positive values
are convex terrain forms the minimum concave extreme values of the first inter-
val -110--12 are found in deep valleys their score is 1, the highest score went to
the interval of plane surfaces.
Slopes are a very important index in parcel evaluation, real estate expert
and contractors inspect the slope as a first stage in their studies and reports,
more the slope is high more the terrain is not suitable and the price is low.
Slope was calculated with the ArcMap algorithm based on DSM and clas-
sified manually in table 8.9, the slope of the study area is moderate 83% of the
area fall in the slope interval between 0 and 29 percent, high score went to the
lowest interval 0-14 and the lower score of one is for the extreme slopes.

80
 Table 8.9: Percent Slope Ranges of DSM Data Sets, Associated with Scores.

Slope
Score Interval % Area %
5 0-14 43
4 15-29 40
3 30-44 12
2 45-59 4
1 > 60 1

Fig. 8.11: Сriteria classified maps a) local relief, b) slope c) curvature, d) resulted real estate.

The resulted three classified datasets maps of local relief, slope and curva-
ture in figure 8.11 shows the high score values in red color and the low score
pixels in dark green.
The method used is weighted overlay analysis, performed by overlaying
classified datasets, assigning a weight to each dataset, summing the values of
each vertical cell stack, and then evaluating the resulting composite map (Col-
lins et al., 2001).
The developed composite map of parcel evaluation for real estate requires
the analysis of criteria illustrated in Table 8.9, in our study, criteria are not all of
equal importance. In a single criteria map we must prioritize values. Values have
been reclassified in the input criteria maps, slope has the big influence on terrain

81
evaluation and assessments that’s why it took a weight of 60% and for Local
relief and curvature 20 % for each. Criteria maps have been weighted and ag-
gregated together to produce parcel evaluation composite map figure 8.11d.
High scores in composite map figure 8.11d are found in the smoothed
plain areas hence low scores are found in extreme valleys. The cadastral vector
map of the study area draped on the resulted raster composite map with zonal
statistics (maximum, minimum and mean) of the scores inside parcels bounda-
ries, the evaluation real estate map classified manually in 5 classes using the
mean values within each parcel.

Fig. 8.12: Parcels evaluation real estate map.

The parcels evaluation real estate map of the study area constitutes a real
estate ranking coefficient, parcels of the first class are not influencing on the
square meter price but the class number five is five times better then class one.
The results of this study is satisfactory, it would suffice to have all reliable
and necessary data relating to the case study if better outcomes were to be ex-
pected. These data are easily introduced in the system and can be updated in any
time, we can add more criteria such as neighborhood analysis, accessibility
analysis etc. Therefore, using the developed system, results should be closer to
reality since all criteria and them relative importance is taken into account at the
same time.
The main aim of this research was to determine whether UAVs DSM can
offer a suitable material for real estate parcels evaluation. To achieve that, a
UAV flight was done to produce DSM for terrain analysis. Afterwards the ter-

82
rain analysis results provide parcels evaluation criteria used for the creation of a
parcel evaluation real estate map.
The photogrammetry processing results in terms of precision are accepta-
ble, since the level of precision only depends on pixel size.
The resulted values of figure 8.12 could be a coefficient of parcel pricing
and real estate market arrangement, the parcel evaluation real estate map is a
basic for decision makers and experts.
The issue addressed is land evaluation for real estate. The ideal solution
would be to incorporate a module including important classification methods in
a GIS as well as appropriate analysis methods independently of data, of the
study area. It will represent a spatial decision support system dedicated to devel-
oping land evaluation map for real estate parcels assessments.

Multi-scale Terrain Roughness

Surface roughness could be defined as a value ranging between smooth
and complex surfaces, this study specifically focuses upon the broad area at dif-
ferent scales of general geomorphology (Evans, 1729) and, more explicitly, on
the quantification of surface-roughness variability using Digital Surface Models
(DSMs) generated from UAV. Surface roughness is treated here as a geomor-
phometric variable influencing at the physiography of the terrain, not as a pa-
rameter due to the precision and accuracy of the generated digital surface mod-
els.
Measurement of terrain roughness is important for a number of disciplines
of terrain quantifying characteristics have been evolving within fields such as
geomorphology, engineering, biologists and ecologists (Doumit, 2017).
In terrains descriptions, roughness parameters should be established that
can be used to describe surface irregularities and they should fulfill some re-
quirements. The parameters should be descriptive and give the reader an image
of the physical characteristics of the study area and should be easily measurable
in the field so that large sites can be quickly sampled. If possible, roughness pa-
rameters should be selected that require similar types of field measurements
with a minimal amount of equipment. Nowadays with the appearance of the
Unmanned aerial vehicles and the advanced of Geographical Information Sys-
tems these parameters can be measured and compared at several different scales,
and suitable for statistical and numerical analysis.
The simplest traditional method of terrain complexity is the profile meth-
od, by providing multi-sections on the terrain it is very easy to evaluate rough-
ness of the terrain.
Hobson among the first scientists who calculated terrain Roughness using
computer technologies, he wrote in Fortran language modules for calculating
roughness parameters such as: comparison of the estimated actual surface area
with the corresponding planar area; bump elevation frequency distribution; and
the distribution of planes (Hobson, 1972).

83
 With the fast evolution of GIS and geoinformatics methods, many scien-
tists worked on the development of other methods for calculating terrain rough-
ness such as: the application of Fourier analysis (Stone and Dugundji, 1965) ge-
ostatistics (Herzfeld, et.al., 2000), the fractal dimension of a surface (Elliot,
1989; Doumit, Pogorelov, 2017).
From the first recognized traditional methods for quantifying roughness
was the land surface roughness index (LSRI) developed by (Beasom et
al.,1983). This index is a function of the total length of topographic contour lines
in a given area.
(Riley et al., 1999) developed a terrain roughness index (TRI) that is de-
rived digital elevation models (DEM) implemented in a geographical infor-
mation system (GIS). TRI uses the sum of changes in elevation within an area as
an index of terrain roughness.
Based on (Hobson, 1972) method developed for measuring surface
roughness in geomorphology, a Vector Roughness Measure (VRM) quantifies
terrain roughness by measuring the dispersion of vectors orthogonal to the ter-
rain surface.
In this study we tested the regression between VRM and TRI values at the
six different levels and we provided a correlation analysis between the raster
datasets of VRM, and TRI, to examine their distributions within each scale, we
generated scatterplots and calculated descriptive statistics (Min, Max, SD,
skewness, kurtosis and r2) to characterize terrain heterogeneity at different level.
Our study is independent from DSM accuracy and precision it will test
roughness at six different levels expressed by flight height of a drone at 20,
40,60,120,240 and 360 meters. The flight datum was calculated from the same
takeoff points of the drone of the six flights.
As this study is restricted to evaluating array-based geomorphometric
methods for calculating surface Roughness, an input DSM is required for further
analysis. DSM selection criteria were based on spatial resolution, with a high-
spatial-resolution DSM required in order to test the heterogeneity across a range
of resolutions and within the study area presenting multiscale Roughness fea-
tures.
The Terrain Roughness Index (TRI) based on an index described by (Ri-
ley, et. al. 1999) that calculated the sum change in elevation between a grid cell
and its eight neighboring grid cells table 8.10 by squaring the eight differences
in elevation, summing the squared differences, and taking the square root of the
sum. (Valentine,et.al. 2004) calculated the average of the absolute values of the
eight differences in elevation, by using the TRI equation given as:

0,0 1, 1 0,0 0, 1
0,0 1, 1 0,0 1,0 0,0
1,1 0,0 0,1 0,0 1,0
0,0 1,1 /8 (8.4)

84
 Table 8.10: 3×3 grid of the TRI equation values.

1,-1 0,-1 1,-1

-1,0 0,0 1,0
-1,1 0,1 1,1

Fig. 8.13: TRI maps at different flight altitudes 20, 40, 60,120,240 and 360 above the datum.

TRI high values at FH-20 shows details in ridges and water erosion traces,
in FH-60 structures are very smoothed, FH-120 shows the pixel's boundaries
and at FH-360 the map is totally pixelated. It is very clear in this map the disap-
pearance of the small structures with the loss of spatial resolution, running from
coarse to smooth then to pixelated surfaces.
Based on a method developed for measuring surface roughness in geo-
morphology (Hobson, 1972), the surface of elevation values can be divided into
planar triangles very similar to Triangulated Irregular Network (TIN models)
and normal to these planes represented by unit vectors. Values of vector mean
strength (R), and dispersion (k) can be calculated for each square cell. In smooth
areas, with similar elevations, the vector strength is expected to be high and the
vector dispersion to be low since the vectors will become parallel figure 8.14. In
rough areas, the nonsystematic variation in elevation will result in low vector
strength and high vector dispersion. The inverse of k can be a better representa-
tion of roughness (Mark, 1975).

85
 Based on slope and aspect definitions, normal unit vectors of every grid
cell of a digital elevation model (DEM) are decomposed into x, y and z compo-
nents.

Fig. 8.14: Vector dispersion method used to calculate surface roughness at different scales for
a topographical surface. Graphic from (Grohmann et al. 2011).

DSM resolution dependent from the flight height, in figure 8.14 the topo-
graphic surface profile showing the terrain variation, at high spatial resolution
vectors are very dense and orientated in several directions otherwise for low spa-
tial DSM resolution as per example FH-360 vectors and far from each other per-
pendicular to segments expressing geometrical terrain forms.
The translation from the vector dispersion traditional method applied on
topographic maps to Vector Roughness Measure (VRM) calculated by GIS algo-
rithms, was done by applying the method and formulas used by (Veitinger et al.,
2016). Based on slope and aspect definition, the normal unit vector of every grid
cell of a Digital Surface Model is decomposed into x, y, and z.
A resultant vector R is then obtained for every pixel by summing up the
single components of the center pixel and its neighbors using a moving window
technique.
∑ ∑ ∑ (8.5)

The magnitude of the resultant vector is then normalized by the number of

grid cell and subtracted from 1.
1 (8.6)
Where VRM is the vector ruggedness measure (Veitinger et al., 2016).
Figure 8.15 shows the six VRM maps generated from DSM, by using
formulae 6, for the first three high spatial resolution FH-20, FH- 40 and FH-60,
terrain structure are very fine highlighted similar to TRI map of figure 8.15. The
two indices TRI and VRM of the resulted roughness maps showed a loss in ter-

86
rain heterogeneity and a trend to terrain homogeneity by a high degree of
smoothness especially in the last three DSMs FH-120, FH-240, and FH-360.
VRM measures the variation in terrain independent of its overall gradient, VRM
is able to differentiate among terrain types.

Fig. 8.15: VRM maps of the six DSMs

In this work, we have tested two widely used methods: Terrain Roughness
Index (TRI), Vector Roughness Measure (VRM), Terrain Roughness Index
(TRI) calculates the sum change in elevation between a grid cell and its neigh-
borhood, according to the algorithm by (Valentine, et.al., 2004).

Table 8.11: Terrain Ruggedness Index statistical values at each level. Std. – standard devia-
tion; Skew – Skewness; n – number of cell in a raster grid.

Mean Std Skew. Kurtosis n Min. Max. Median r2

TRI-20 0.116 0.062 1.202 2.861 40436 0.005 0.544 0.112 0.0014
TRI-40 0.171 0.085 0.702 1.233 18522 0.008 0.591 0.172 0.00006
TRI-60 0.211 0.101 0.602 0.575 8891 0.016 0.636 0.208 0.0059
TRI-120 0.520 0.198 -0.424 -0.045 1901 0.037 1.093 0.552 0.0081
TRI-240 0.822 0.302 -0.385 -0.323 559 0.136 1.695 0.879 0.0033
TRI-360 1.113 0.381 -0.723 -0.448 286 0.152 1.874 1.234 0.0292

The statistics of the TRI values at each flight height listed in table 8.11,
the values of Min., Max., Mean and Std. showed that the TRI values increased
with the flight height hence with the scale. From the values of r2 it is proven that
no homogeneity of TRI values with their neighborhoods in each layer, it is nor-

87
mal especially for the high spatial resolution layer TRI-20, TRI-40, and TRI-60
with high n values.
For TRI-20 no symmetric data distribution because of the high skewness
value of 1.202, but the evidence is that negative values for the skewness at TRI-
120, TRI-240 and TRI-360 indicate data that are skewed left and positive values
for the skewness indicate that high spatial resolutions layer TRI-20, TRI-40, and
TRI-60 skewed right.

Table 8.12: Vector Ruggedness Measure statistical values at each level.

Mean Std Skew. Kurtosis n Min. Max. Median r2

VRM-20 0.021 0.019 2.219 6.906 40436 0 0.166 0.017 0.013
VRM-40 0.021 0.017 1.753 4.757 18522 0 0.130 0.018 0.007
VRM-60 0.015 0.012 1.512 3.137 8891 0.0001 0.080 0.013 0.026
VRM-120 0.019 0.011 0.384 0.563 1901 0.0001 0.063 0.020 0.0001
VRM-240 0.015 0.008 0.119 -0.561 559 0.0006 0.037 0.015 0.0009
VRM-360 0.014 0.006 -0.289 -0.914 286 0.0009 0.027 0.015 0.0081

The distributions of roughness values (VRM) for the five levels were
highly skewed to the right with the highest proportion of VRM values at the
mean instead of FH-360 values skewed to the left.
Our results showed that TRI and VRM directly measured heterogeneity of
terrain more independently of scale, and both indices exhibited a pattern of bias
in that the minimum value of roughness increased with increasing spatial resolu-
tion.
A correlation analysis provided to understand the similarity between TRI
and VRM.

Fig. 8.16: Scatterplot of TRI and VRM ruggedness values at all levels of details. a) FH-20,

88
 b) FH-40, c) FH-60, d) FH-120, e) FH-240, f) FH-360.

High correlation recorder at all flight heights, the scattered plot of figure
8.16 shows a high degree of similarity in small values at FH-20, FH-40 and FH-
60 expressed in dark color elongated areas of figures 8.16 a b and c.
At high flight height the concentration of the correlated values is moving
from small to mean values with a trend to the right figure 8.16e, otherwise the
correlation values of TRI and VRM in figure 8.16 f became more scattered and
less dense due to a dilution of similarity resulted from the changing of the spatial
resolution (pixel size).
We can say from figure 8.16 that the two roughness indices are very simi-
lar and have a high correlation and the degree of terrain roughness vary with the
spatial resolution. Differences in the distributions of roughness, measured by
VRM, and TRI reflected the characteristic terrain physiography of the terrain.
Surface Roughness in Earth sciences is used as an explanatory index. It is
dependent upon exogenic and endogenic geographical processes. Many methods
for surface Roughness measuring such as: area ratio, vector dispersion, the
standard deviation of first and second terrain derivative (elevation, slope, and
curvature) have been implemented in GIS and based on digital models.
The possibility of the production of digital models at different spatial
resolution spatially UAV based one, allows fast and inexpensive multiscale
analysis of surface Roughness. Two applied indices Topographic Roughness
Index (TRI) and Vector Roughness Measure (VRM) at different scale level ex-
press a variety in terrain heterogeneity at a UAV flight height of 20, 40, 60, 120
240 and 360.
Both indices show a roughness variation with scales and a transition from
coarse to smooth between FH-60 and FH-120, a cartographic generalization in-
fluenced by flight height is very clear in figure 8.14 and 8.16. Our statistical and
correlation analysis of roughness indices prove that multiscale and multilevel
UAV flights datasets are: a visual cartographic generalization, a transition scale
from level to another, a live roughness monitoring apparatus leads to a detection
of fine scale/regional relief, and performance at a variety of scales.
Researchers must be aware of potential biases that originate in DSM at
multiscale (different spatial resolution) when TRI and VRM values are inter-
preted. All DSMs contain inherent inaccuracies due to the sources errors in orig-
inal data. The elevation accuracy of a DSM is greatest in flat terrain and de-
creases in steep terrain where the roughness incises (Koeln et al., 1996). Terrain
roughness is a complicated geomorphometric parameter, it could be calculated
in many ways, under many names roughness, micro relief, and others.

Digital Surface Models to planar areas.

Rugosity is a measure of topographic heterogeneity, it describes and
counts discrete structures, from categorical to quantitative (McCormick, 1994).

89
Rugosity is an index of surface roughness that is widely used as a measure of
landscape structural complexity.
Rugosity is traditionally evaluated in-situ across a two dimensional terrain
profile by draping a chain over the surface and comparing the length of the chain
with the linear length of the profile figure 8.17.

Fig. 8.17: The rugosity of a surface (e.g. yellow profile of a terrain № 1) is the ratio
between the contoured distance (dashed line № 2) and the planar distance
(or area for three-dimensional data).

Two-dimensional rugosity is defined as the ratio between the surface con-

tour distance and the linear distance between two points (Risk, 1972) and is syn-
onymous with the term tortuosity or the arc-chord ratio (Moser et al., 2007).
There are multiple measures of structural complexity (McCormick, 1994),
for over forty years many scientists have used the rugosity index for topographic
(elevation) or bathymetric (depth) datasets; rugosity can be calculated from two-
or three-dimensional data at any scale.
Drones based Digital Surface Models at different flight heights with dif-
ferent scale could be a good material for rugosity analysis at different scales.
The standard surface ratio (SR) method for calculating a planar distance is
to project the surface onto a horizontal plane (solid line), this method confounds
rugosity with the slope (θ) at the scale of the surface (solid line).
Figure 8.17 is a very expressive section, the length of the section number
1 natural terrain with 5.8 km a very complex terrain, section number 2 smoothed
terrain owning a similar shape as the natural terrain but with 5.3 km length. Sec-
tion number three express a slope projection of the natural terrain with a 3.5 km
length.
The last section a horizontal planar one number 4 with 2.8 km influenced
by the slope angle (θ), the big change in length from section 1 to the horizontal
section 4 is approximately the half. The idea here is not the changing in length
nor the surface, but the rugosity change at different DSM spatial resolution.
A six DSM generated from different UAV flight height constitute the base
of our study by comparing the transition to planar areas at different levels.

90
 In the early 1970s The standard surface ratio (SR) method for measuring
rugosity was introduced by (Risk 1972; Dahl, 1973). They calculated rugosity
by projecting the surface onto a horizontal plane (Lundblad et al., 2006; Wright,
Heyman, 2008; Friedman et al., 2012) thereby coupling rugosity with the slope
at the scale of the surface equation 1.

(8.7)

in the case of equation 7 the rugosity increases with increasing of the

slope figure 8.17; the law of cosines, where cos (θ) = adjacent side ÷ hypotenuse
side), here lies the fundamental issue presented by the traditional methods for
measuring rugosity.
A flat surface has a rugosity value of one, while a rougher surface, or a
surface with more relief, has a higher rugosity value than one, rugosity encom-
passes and combines both structural relief and roughness (Moser et al., 2007).
A new arc–chord ratio (ACR) rugosity index for quantifying landscape
structural complexity was developed by Du Preez (2014) to overcome signifi-
cant issues presented by traditional rugosity indices. In comparison to other
methods for measuring rugosity, ACR rugosity is separated from the slope and
easy to execute an ACR rugosity analyses using the GIS software (Du Preez,
2014).
Many marine and land studies use rugosity such as environmental risk as-
sessment, species management, distribution and conservation, predictive map-
ping of vulnerable marine and land ecosystems (Stambaugh, Guyette 2008;
Wedding et al., 2008; Galparsoro et al., 2009; Woodby et al., 2009).
The ACR three dimensional method is tested on the six generated DSM
and compared at different levels of spatial resolution by correlation and statisti-
cal analyst.
Spatial resolution is a very important factor of scale influence on rugosity,
to prove that our study will find the answer to the question: did the transition of
the multiscale Digital Surface Models to planar areas have the same results?
Many methods of evaluating rugosity on a three dimensional surface have
been proposed. These methods measure a ratio of areas rather than lengths, as
shown in Equation (7).
Surface Area to Planar Area (SAPA) method introduced by Jeff Jenness
evaluates rugosity using a 3 x 3 neighborhood, by drawing a line from the center
of each cell in the window to the center of the central cell in three dimensions.
The result is a network of eight triangles in the central cell which approximates
the contoured surface at the cell location. The sum area of these triangles is di-
vided by the two dimensional cell area to obtain a measure of rugosity (Jenness
2004).
Du Preez and Tunnicliffe (2012) propose a novel method for measuring
rugosity that decouples rugosity from the slope and is consistently independent
of data dimensionality and scale. it is a simple adaptation of an Arc Chord Ratio

91
(ACR). The method replaces the horizontal plane with a plane of best fit
(POBF), where the POBF is a function of boundary data interpolation (Preez,
Tunnicliffe, 2012). The ACR method can be used in multi-scale analyses, an
important attribute of a spatial analysis as morphological processes act at a vari-
ety of spatial scales (Levin, 1992), and differ in effects and importance with
scale (Wu, 2013).
Basing on Du Preez (2014), Jeff Jenness developed A new technique, op-
erates on a 3×3 neighborhood, using the triangulated area of each adjacent cell
and applying the Pythagorean theorem to compute the surface area. By default,
the planar area of each grid cell is corrected by dividing the cell area by the co-
sine of slope (Jenness, 2004).
In our study ACR was calculate in GIS tool installed on ArcGIS® Soft-
ware available for download in Du Preez (2014), from the six generated DSMs
we calculated Arc-Chord ration (ACR). The ACR rugosity index is a measure of
three-dimensional structural complexity defined as the contoured area of the sur-
face divided by the area of the surface orthogonally projected onto a plane of
best fit.
The arc-chord ratio (ACR) method calculates the planar distance by pro-
jecting the surface boundary onto a boundary data section 1 figure 8.17 (Red
dashed line) plane of best fit section 3 figure 8.17 (POBF; dashed-dotted line; 3)
effectively decoupling rugosity from the slope at the scale of the surface. ACR is
calculated by creating two TINs a contoured surface and a planar one represent-
ing the plane of best fit (POBF) figure 8.18a. The POBF is a function (interpola-
tion) of the boundary data only of the area of interest the area of interest is the
boundary of the study area. The surface area of the first TIN within is divided by
that of the second TIN to obtain a single ACR value for the area of interest (Du
Preez, 2014).

Fig. 8.18: A) ACR simultaneous surfaces leading to the horizontal planar one, b) an example
of ACR surfaces of FH-20.

As a first step, the conversation from raster to TIN for the six DSMs to
form contour surfaces at different scales.

92
 Fig. 8.19: TIN models of contour surface at the six flight heights.

All the six TIN models expressed the terrain morphology with some varia-
tions detected on the colored contour lines, and a generalization in triangles
quantity.
Step two, the contoured surface translated to a plane of best fit (POBF)
decoupling rugosity from the slope at the scale of the surface (Du Preez, Tunni-
cliffe, 2012; Friedman et al., 2012).
Figure 8.20 shows six similar planar surfaces owning the same trend of
values by simplifying elevation values of the contour surfaces.
The innovation of the ACR method lies in the analysis used to generate
the POBF: identify and isolate the boundary data (step three) figure 8.18a an
illustrated example the boundary data of FH-20 the triangulated irregular net-
work data frame.
By using a linear polynomial interpolation of the boundary data at the six
levels to generate surface datasets (step four).

93
 Fig. 8.20: the POBF of the six flight heights.

Some softwares are unable to interpolate the actual planar area, an alterna-
tive is to interpolate the angle of the POBF and use the cosine equation (and the
horizontal planar area) to extract the planar area of step five (Du Preez 2014).
To solve for the ACR rugosity index, by application of formula 7 using
the contoured and planar areas (step six).
By following Du Preez and Tunnicliffe (2012) arc-chord ratio (ACR) we
Computed a ratio between the three-dimensional surface area and the planar area
of the surface, this tool uses a novel methodology to develop a surface area da-
taset. The output values represent ratios between the surface area and planar ar-
ea, typically ranging from 1 in flat areas to 4 in areas of high variation.
The first step of Du Preez methodology of conversation from raster to
TIN, figure 8.19 of the similar six TIN models. Same study area at different spa-
tial resolution lead to visual data similarity a statistical comparison was done to
test this degree of similarity table 8.13.
The statistical values of table 8.13 shows a decreasing in triangles num-
bers from 321 to 41 due to decreasing in spatial resolution, the difference in av-
erage maximum and minimum elevations due to the interpolated predicted val-
ues.

94
 Table 8.13: Elevation TIN statistics, quantity of triangles, average minimum and maximum
elevations and the slope average.

Contoured area TIN

Flight height, Quantity of Average Min. Average Max. Average
Meter triangles Elev., Meter Elev., Meter Slope %
20 321 1728.95 1729.94 15.89
40 230 1730.26 1731.53 16.70
60 172 1733.58 1734.85 14.52
120 85 1718.65 1720.76 17.03
240 58 1727.14 1729.25 14.48
360 41 1746.67 1748.87 13.00

The quantity of triangle from elevation area to planar area of table 8.13
and table 8.14 are reduced more than twenty times in high spatial resolution data
of 20, 40 and 60 meters’ flight heights, otherwise low spatial resolution data
with low triangles quantity in contoured areas are reduced less than ten times in
planar areas TIN models.
Average of maximum and minimum elevations approximately in all levels
are reduced in a range of 2 meters between surface and planar areas, hence a re-
duction on the average slope in the range of 2 percent.
Table 8.14: POBF, planar TIN area statistics quantity of triangles, average minimum and
maximum elevations and the slope average.

Planar area TIN

Flight height, Quantity of Average Min. Average Max. Average
Meter triangles Elev., Meter Elev., Meter Slope %
20 10 1725.72 1731.67 13.23
40 9 1726.72 1732.41 14.13
60 9 1732.90 1732.98 10.44
120 8 1715.46 1722.23 14.87
240 8 1724.62 1729.84 11.72
360 7 1746.22 1746.66 10.75

In planar area TIN, the average values of the minimum and maximum el-
evations in all the six flights is reduced with the number of triangles due to the
transition from contoured to planar area.
The variation in values between table 8.13 and table 8.14 showed an un-
stable change in elevations and slope.
The first part of the transition from contoured area TIN to planar area TIN
(POBF) is very similar to trend analysis simplifying the complexity of values
with a conservation of the same datum.
Otherwise the second transition part from surface area two planar one rec-
ord a loss of initial datum elevation down to zero table 8.15.

95
 Table 8.15: Surface areas statistics at different flight heights.

Surface Area (Meter)

Flight heights Min Max Mean Std
20 0.141 0.227 0.144 0.005
40 0.312 0.415 0.319 0.008
60 0.645 0.756 0.657 0.013
120 3.007 3.342 3.073 0.048
240 10.7 11.3 10.867 0.105
360 19.976 20.705 20.291 0.172

High spatial resolution data of 20, 40 and 60 have sub meter surface area
values, rising with approximately in double values between flight heights.
The surface area is designed to determine the amount of similarity be-
tween the tested area surface and planar surface. it is hypothesized that the sur-
face area increases with surface irregularity. Because there is a definite interplay
between the number and magnitude of terrain irregularities such that similar sur-
face area estimates could arise from different manipulations of these two varia-
bles.

Table 8.16: Planar area statistics at different flight heights.

Planar Area (Meter)

Flight height Min Max Mean Std
20 0.99 1.154 1 0.005
40 0.99 1.132 1 0.003
60 0.99 1.045 1 0.002
120 0.99 1.013 1 0.001
240 0.99 1.003 1 0.0006
360 0.99 1.003 1 0.0005

The values of planar area at different scales is practically the same with
small variations at low flight height, basing the results of table 8.16 especially
the similar mean values, we can answer the above question constituting the tar-
get of our study by: yes, the transition of the multiscale Digital Surface Models
to planar areas have the same results.
Visual and statistical results prove the similarity of multiscale planar data,
a regression analysis run to test this similarity in planar surface at multiscale.

96
 Fig. 8.21: scattered plots of planar area at different scales.

The correlation analysis between the highest spatial resolution data of FH-
20 as a reference and the other datasets, figure 8.21a a test scatterplot with same
data set in X and Y axes of FH-20, with hundred percent similarity.
The graph FH-40 with FH-20 gives 39.77 % of similarity figure 8.21b, r
square values of FH-60, FH-120, FH-240 and FH-360 against FH-20 are less
than 11 %, the core of the scatterplots for the high spatial resolution datasets are
in the lower left corners, moving positively with the Y axes in FH-120 then fall-
ing down negatively for FH-360.
The correlation analysis contrary to visual and statistical showed a differ-
ence in multiscale planar areas.
The present study provides multiscale DSM analysis by adaptation and
improvement of ACR geo-processing model tools and step-by-step application
of Du Preez 2014 module. Improving standard methods for the detection and
investigation of geomorphological patterns at different spatial resolution will
lead to better scientific information for generalizations, terrain analysis, man-
agement and conservation initiatives.
We can conclude from this study that Scale and resolutions effects on ter-
rain data and form an important issue in geographic researches, DSM UAV
based at high flights should be tested before use.
It is essential to have a good understanding of the effects of scale on the
analysis results, each elevation data has its own surface to planar result, terrain
rugosity depends with spatial resolution and visual analysis should follow a cor-
relation one.

97
 Chapter 9 Drones regulations and buyers guide

Drones Regulations
When it comes to flying professional drones within the law, the respective
rules around the world are varied, till now there is no unique regulations for all
countries. Aviation authorities around the world are integrating unmanned air-
craft into civilian airspace, and each jurisdiction has its own rules and regula-
tions. As an example in USA, you must acquire a Certificate of Authorization
(COA) through the Federal Aviation Administration (FAA). While the require-
ments of each COA include (Reg Austin, 2010):
 Flights below 300 meters. Daytime operation in Visual Flight Rules
(VFR).
 Range limited to Visual Line of Sight (VLOS).
 Greater than 4 km from an airport.
 Flight altitude ceiling of 120 m above take-off point is common.
 Maximum UAV take-off weight/weight classes (a lighter weight in-
creasingly equates to more flexible usage).
 No, fly zones, such as within several km of an airport, dense urban ar-
eas and military places.
 An operator certificate/license of some description is often required.
 A second drone observer is sometimes required.
The investigations reveal that UAV regulations are subject to national leg-
islation and focus on three key issues:
1) Targeting the regulated use of airspace by UAVs as they pose a serious
danger for manned aircrafts;
2) Setting operational limitations in order to assure appropriate flights.
3) Tackling administrative procedures of flight permissions, pilot licenses
and data collection authorization in order to address public safety and privacy
issues.
Approximately half of all countries do not provide any information re-
garding the use of UAVs for civil applications, some countries have UAV regu-
lations were in other countries the use of UAVs is prohibited (Stöcker et al.,
2017).

Flight safety
A drone operator is responsible for ensuring the safety of every operation,
including the protection of nearby people, animals, property and the environ-
ment in general, he should evaluate weather conditions and choose suitable, safe
take-off and landing locations (Reg Austin, 2010).
For a safe flight drone operator must:
1. Flight within line of sight LOS.
2. Do not interrupt the LOS.

98
 3. Do not fly with wind over 30 km/h.
4. Do not fly in fog.
5. Do not fly in rain.
6. Do not fly long periods (the ESC got hot).
7. Stay away from birds.
The ICAO is an international actor that serves as a collaboration and
communication platform for national civil aviation authorities. They are con-
cerned with fundamental regulatory frameworks at a global scale and provide
information material, Standard and Recommended Practices and Procedures for
Air Navigation Services (ICAO 2011). One important step taken in Riga 2015
was the publication of the Riga Declaration on Remotely Piloted Aircrafts. The
declaration highlights five main principles that should guide the regulatory
framework in Europe:
1) Drones need to be treated as new types of aircraft with proportionate
rules based on the risk of each operation;
2) EU rules for the safe provision of drone services need to be developed
now;
3) Technologies and standards need to be developed for the full integration of
drones in the European airspace; 4) Public acceptance is key to the growth of
drone services;
5) The operator of a drone is responsible for its use (Riga declaration
2005).
In some countries the required distance bounds a vague interpretation and
strictly determines the term visual line-of-sight (VLOS). Some cases further in-
clude extended visual line-of-sight (EVLOS) operations.
Here, the pilot uses an additional observer or remote pilots to keep the
visual contact to the UAV figure 9.1. The US, UK, Italy and South Africa par-
ticularly mention the possibility of EVLOS operations within their UAV regula-
tions. Furthermore, some countries basically allow beyond visual line-of-sight
(BVLOS) flights.
BVLOS flights, which are outside the general permission for the commer-
cial utilization of UAVs, require either special flight conditions or exceptional
approvals (Stöcker, 2017).

99
 Fig. 9.1: Schematic distinction between UAV flight ranges.

Besides this, the majority of the cases demand a pilot certification or a li-
cense. A certificate is usually granted by intermediaries like authorized training
centers or UAV manufacturers and entails a basic practical and theoretical train-
ing of the pilot. In contrast, the procedure and requirements to obtain a UAV
pilot license usually involve sophisticated aeronautical background knowledge
and is issued by national aviation authorities (Stöcker, 2017).

Drones buyers guide

There are many different types of UAVs on the market today. The solu-
tion you choose depends on various factors, such as the type of data that is to be
collected, the safety requirements, and the time and distance the drone must be
capable of flying in a single mission.
Deciding on the type of UAV that is most suited to your project’s needs
and requirements is not always an easy task. Below some guidance on things to
consider and the questions to ask when you decide to purchase a drone.
Questions to ask when choosing a drone and things to consider before
purchasing a UAV:
The payload, how much weight and volume is needed?
Redundancy, for how long and how far is the drone needed to fly?
Safety, where and when am, I flying? Are there special regulations in
place?
Transport (Collapsible, portable, easy-to-carry)?
Software Accessories (ground station with hardware and software for mis-
sion planning)?
User friendliness, How easy is mission planning, start, landing, flight, and
data collection?
Customer service, is maintenance and support available?
Price, What is the total cost of ownership?
Software, what are the post processing software capabilities included?

100
 Conclusion
Today, UAVs can be used as a precise, automated and computer con-
trolled data acquisition and measurement platform, this work deals with the
challenging task: The use of UAV systems as photogrammetric data acquisition
platform.
The main goal of this book is the identification of a generic workflow us-
ing a UAV as a photogrammetric data acquisition platform investigated on a real
application, focusing on the precision and resolution of the generated photo-
grammetric products, like elevation models, orthophoto and textured 3D models.
The main motivation of this book is to generate high-resolution infor-
mation of inaccessible areas, such as maps, orthoimages, general topography,
detailed elevation models, extraction of dangerous obstacles.
The extraction of the terrain, orthoimages and textured 3D models from
UAV-images or other sensor data can be applied to all kinds of hazards, cata-
strophic or environmental disasters, including 3D documentation of the envi-
ronment, cultural heritage sites, etc… Drones technologies allow cartogra-
phers to produce better mapping, coupled with multispectral, hyperspectral and
LiDAR systems.
The evaluation of low-cost UAV systems showed that these systems are
easily controllable and usable for scientific researches. Many challenges still
remain as UAV are adopted in the geospatial industry, this book showed that
flying the photography did not present as many challenges as the post-
processing of the resulting spatial data.
This book demonstrated that unmanned aerial vehicles (UAV) can be used
to produce geospatial data in the geographical analysis. The modular nature of
UAS and the ability to produce current spatial information (orthophotos, 3D
models) mean that we no longer need to turn to large mapping contracts because
of the economies of scale offered by the latter.
Today, UAVs can be used as a precise, automated and computer con-
trolled data acquisition and measurement platform, this work deals with the
challenging task: The use of UAV systems as photogrammetric data acquisition
platform.
The key with using all photography or imagery, whether aerial or satellite-
derived, is to only use the data in a sensible and meaningful way, understand the
error levels and their potential impact. Geographic correction can be a compli-
cated process, full of pitfalls it should transform a better image into a useful GIS
information.
The acquired data has great potential in GIS projects. Moreover, our study
showed also its usability for teaching Students to use new technologies.in future
work UAVs will be integrated as a fixed part in the remote sensing and GIS
courses.

101
 References

1. Abbeel P., Coates, A., Quigley, M. and Ng, A. Y. (2007). An application of reinforce-
ment learning to aerobatic helicopter flight, In: NIPS, 19.
2. Abdel-Aziz Y. I. and Karara H.M. (1971). Direct linear transformation into object space
coordinates in close-range photogrammetry. In Symposium on Close-Range Photo-
grammetry.
3. Adams L.P. (1980). The Use of Non Metric Cameras in Short Range Photogrammetry.
14th Congress of the International Society for Photogrammetry, Commission V, Ham-
burg, Germany.
4. Agisoft, (2013). Agisoft PhotoScan User Manual: Professional Edition. Version 1.0.0
edn.
5. Agüera-Vega F., Carvajal-Ramírez, F., Martínez-Carricondo, P. (2016).: Accuracy of
digital surface models and orthophotos derived from unmanned aerial vehicle photo-
grammetry. J. Surveying Eng. 04016025. and Pattern Recognition, 1997.
6. archaeological areas using a low-cost UAV the Augusta Bagiennorum Test site, In: XXI
7. Arrell K, Fisher P.F, Tate N.J, Bastin L (2007) A fuzzy c-means classification of eleva-
tion derivatives to extract the morphometric classification of landforms in Snowdonia,
Wales. Computers & Geosciences 33, pp.1366-1381.
8. Atkinson K. B. (1996). Close Range Photogrammetry and Machine Vision. Whittles
Publishing, Roseleigh House, Latheronwheel, Caithness, Scotland.
9. Austin Reg. (2010) Unmanned aircraft systems UAVs design development and deploy-
ment, John Wiley & Sons Ltd, p.332
10. Autodesk. (2014). 123D Catch.www.123dapp.com/catch.
11. Barka I., Vladovic J., Malis F. (2011) Landform classification and its application in
predictive mapping of soil and forest units. Proceedings: GIS Ostrava 2011.
12. Barnes G., Volkman W., Sherko R., Kelm K. (2014). Design and Testing of a UAV-
based Cadastral Surveying and Mapping Methodology in Albania. World bank conf.
Land and Poverty, The World Bank - Washington DC, March 24-27.
13. Beasom S.L., Wiggers E.P., Giordono R. J. (1983). A technique for assessing land sur-
face ruggedness. Journal of Wildlife Management. 47: pp. 1163–1166.
14. Bellingham J.S., Tillerson M., Alighanbari M. and How J. P. (2003). Cooperative Path
Planning for Multiple UAVs in Dynamic and Uncertain Environments, In: 42nd IEEE
Conference on Decision and Control, Maui, Hawaii (USA).
15. Bendea H. F., Chiabrando F., Tonolo F.G. and Marenchino D. (2007). Mapping of ar-
chaeological areas using a low-cost UAV the Augusta Bagiennorum Test site, In: XXI
International Symposium, Athens, Greece.
16. Boufama B., Mohr R., Veillon F. (1993) Euclidean constraints on uncalibrated recon-
struction. Proceeding of the Fourth International Conference on Computer Vision, Ber-
lin, Germany, 466-470.
17. Brown D.C. (1976). The bundle adjustment—progress and prospects. Int. Archives Pho-
togrammetry, 21(3), Paper number 3–03 (33 pages).
18. Burrough P. A., McDonell R.A. (1998). Principles of Geographic Information Systems.
Spatial Information and Geostatistics, Chapter 8: Spatial Analysis Using Continuous
Fields, ISBN 0-19-823365-5.
19. Burrough P.A. (1983). Multiscale sources of spatial variation in soil- the application of
fractal concepts to nested levels of soil variation, J. Soil Sci., 34, 577.
20. Chang K., Tsai B (1991) The effect of DEM resolution on slope and aspect mapping.
Cartography and Geographic Information Science 18, pp. 69–77 scheduling, Whistler,
British Columbia, Canada.

102
21. Collins M. G., Steiner, F. R., & Rushman, M. J. (2001). Land-Use Suitability Analysis
in the United States: Historical Development and Promising Technological Achieve-
ments. Environmental Management, 611-621.decision-making methods. International
Journal of Geographic Information.
22. Cooper M.A. R. and Cross P.A. (1991). Statistical concepts and their application in pho-
togrammetry and surveying. Photogrammetric Record, 13(77):645–678.
23. Dahl A.L. (1973). Surface area in ecological analysis: quantification of benthic coral-
reef algae. Mar Biol. 23, pp.239–249.
24. Dai F., & Lu M. (2010). Assessing the accuracy of applying photogrammetry to take
geometric measurements on building products. Journal of Construction Engineering and
Management, 136(2), 242–250. Reston, VA: ASCE.
25. Dellaert F., Seitz S.M., Thorpe C.E. & Thrun, S. (2000) Structure from motion without
correspondence. Proceedings IEEE Conference on Computer Vision and Pattern
Recognition. CVPR 2000 (Cat. No.PR00662). pp. 557-564. IEEE Comput. Soc, Hilton
Head Island, SC, USA.
26. Deng Y., Wilson J.P., Bauer B.O. (2007) DEM resolution dependencies of terrain at-
tributes across a landscape. International Journal of Geographical Information Science
21, pp. 187–213.
27. Dissanayake G., Sukarieh S., Nebto E. and Durrant Whyte H. (2001) “The Aiding of a
Low-Cost Strap Down Inertial Measurement Unit Using Vehicle Model Constraints for
Land Vehicle Applications”, IEEE transactions on robotics and automation, Vol. 17.
28. Doumit J.A., Kiselev E.N. (2016). Structure from motion technology for macro scale ob-
jects cartography// Breakthrough scientific research as the modern engine of sciences,
St. Petersburg.: Publishers “Cult Inform Press”. pp 42-47. ISBN 978-5-8392-0627-4.
29. Doumit J.A., Pogorelov A.V. (2017). Multi-scale Analysis of Digital Surface Models
Based on UAV Datasets. Modern Environmental Science and Engineering (ISSN 2333-
2581), Volume 3, No. 7, pp. 460-468.
30. Du Preez, C. (2014). "A new arc-chord ratio (ACR) rugosity index for quantifying
three-dimensional landscape structural complexity." Landscape Ecology. 30, pp. 181–
192.
31. Du Preez C., Tunnicliffe V. Shortspine thornyhead and rockfish (Scorpaenidae). (2011).
distribution in response to substratum, biogenic structures and trawling. Mar. Ecol.
Prog. Ser. 425, pp. 217-231. [doi:10.3354/meps09005]. dynamic detection of moving
vehicles with a UAV, In: IEEE International Conference on Robotics and Automation,
pp. 1878-1883.
32. Eastman J.R., (1985). Single-Pass Measurement of the Fractional Dimensionality of
Digitized Cartographic Lines. paper presented to the Canadian Cartographic Associa-
tion, Annual Meeting, June 1985.
33. Egbert J. and Beard, R. W. (2007). Road following control constraints for low altitude
miniature air vehicles, In: American Control Conference, New York, U.S., 353-358.
34. Eisenbeiss H. (2009). UAV photogrammetry. Dissertation ETH No. 18515, Institute of
geodes and Photogrammetry, ETH Zurich, Switzerland, doi:10.3929/ethz-a-005939264.
35. Eisenbeiss H. (2011). The Potential of Unmanned Aerial Vehicles for Mapping, Zurich.
Photogrammetric Week '11 Dieter Fritsch (Ed.) Wichmann/VDE Verlag, Belin & Of-
fenbach.pp135-144.
36. Elliot J. K. (1989). An investigation of the change in surface roughness through time on
the foreland of Austre Okstindbreen, North Norway,” Comput. Geosci., vol. 15, no. 2,
pp. 209–217.

103
37. Evans I. (2003) Scale-specific landforms and aspects of the land surface. In: Evans, I.S.,
Dikau R., Tokunaga E, Ohmori H, Hirano M. (Eds.), Concepts and Modelling in Geo-
morphology: International Perspectives. Terrapub, Tokyo, pp. 61–84.
38. Evans S. (1990) “General Geomorphometry”. In: Goudie, A.S., Anderson M., Burt T.,
Lewin J., Richards, Whalley K., Worsley B., Geomorphological Techniques. 2nd edi-
tion. Unwin Hyman, London, pp.44–56.
39. Evans I.S. (1972) General geomorphometry, derivatives of altitude, and descriptive sta-
tistics. In: Chorley, R.J. (Ed.), Spatial Analysis in Geomorphology. Methuen, London,
pp. 17–90.
40. Falkner E., Morgan D. (2002) Aerial Mapping: Methods and Applications, (Mapping
Science) 2nd Edition, by CRC Press LLC, pp.195. ISBN 1-56670-557-6.
41. Faugeras O. D., Luong Q., and Maybank S. 1992 Camera self-calibration: Theory and
experiments. In Proc. European Conference on Computer Vision, LNCS 588, pages
321–334.
42. Florinsky I.V., Kuryakova G.A. (2000) Determination of grid size for digital terrain
modelling in landscape investigations-exemplified by soil moisture distribution at a mi-
cro-scale. International Journal of Geographical Information Science 14, pp. 815–832.
43. Friedman A., Pizarro O., Williams S.B. (2012). Johnson-Roberson M. Multi-scale
measures of rugosity, slope and aspect from benthic stereo image reconstructions. PLoS
One 7, 12, pp. 1–14.
44. Furukawa Y. & Ponce, J. (2009) Accurate, Dense, and Robust Multi-View Stereopsis.
IEEE Transactions on Pattern Analysis and Machine Intelligence. p. IEEE Computer
Society Press.
45. Galparsoro I., Borja A., Bald J., Liria P. (2009). Chust G. Predicting suitable habitat for
the European lobster (Homarus gammarus), on the Basque continental shelf (Bay of
Biscay), using Ecological-Niche Factor analysis. Ecol Model 220, pp. 556–567.
46. Glover J.M. (2014). Drone University, USA, Middletown, DE. P.134.
47. Gonzalez, J. P., Nagy, B. and Stentz, A., (2006). The Geometric Path Planner for Navi-
gating Unmanned Vehicles in Dynamic Environments, In: Proceedings ANS 1st Joint
Emergency Preparedness and Response and Robotic and Remote Systems, Salt Lake
City (Utah), USA.
48. Gonzalez M.F. and Wintz, P. (1987). Digital Image Processing, Addison Wesley,
Menlo Park, CA, p.414.
49. Goodchild M.F. and Mark D.M. (1987). The fractal nature of geographic phenomena.
Ann. Assoc. Am. Geogr., 77, p.265.
50. Granshaw S. (1980). Bundle adjustment methods in engineering photogrammetry. Pho-
togrammetric Record, 10(56), pp.181–207.
51. Grohmann C., Smith M., Riccomini C. (2011). Multiscale analysis of topographic sur-
face roughness in the Midland Valley, Scotland, IEEE T. Geosci. Remote, 49, pp.
1200–1213.
52. Gruner H. (1977). “Photogrammetry: 1776-1976”, Photogrammetric Engineering and
Remote Sensing, 43(5), pp.569-574.
53. Haarbrink, R. B. and Koers, E. (2006). Helicopter UAV for photogrammetry and rapid
response, In: International Archives of Photogrammetry, Remote Sensing and Spatial
Information Sciences, ISPRS Workshop of Inter-Commission WG I/V, Autonomous
Navigation, Antwerp, Belgium.
54. Hartley R.I., Zisserman, A. (2003): Multiple View Geometry in Computer Vision.
Cambridge University Press

104
55. Hengl T., Evans, I.S. (2009). Mathematical and digital models of the land surface. In:
Hengl T., Reuter, H.I. (Eds.), Geomorphometry — Concepts, Software, Applications.
Developments in Soil Science, vol. 33. Elsevier, Amsterdam, pp. 31–63.
56. Hengl T. (2006) Finding the right pixel size. Computers & Geosciences 32, pp. 1283–
1298.
57. Herzfeld U.C., Mayer H., Feller W., Mimler M. (2000) Geostatistical analysis of glaci-
er-roughness data, Ann. Glaciol., vol. 30, no. 1, pp. 235–242.
58. Heyden A. and Astrom K. (1997). Euclidean reconstruction from image sequences with
varying and unknown focal length and principal point. In Proc. IEEE Conference on
Computer Vision and Pattern Recognition.
59. Hobson R. D. (1972). Surface roughness in topography: quantitative approach. in R. J.
Chorley, editor. Spatial analysis in geomorphology. Harper and Row, New York, New
York, USA, pp. 221–245.
60. Hongxia C., Zongjian L., Guozhong S. (2007). Non-metric CCD Camera Calibration for
Low Attitude Photogrammetric Mapping. The Eighth International Conference on Elec-
tronic Measurement and Instruments ICEMI’.
61. Hopkins, L. (1977). Methods for Generating Land Suitability Maps: A Comparative
Evaluation. Journal of the American Institute of Planners, 380-400.
62. ICAO, (2011). Unmanned Aircraft Systems (UAS). Published in separate Arabic, Chi-
nese, English, French, Russian and Spanish editions by the International civil aviation
organization. University Street, Montréal, Quebec, Canada H3C 5H7. P. 38. ISBN 978-
92-9231-751-5
63. Irschara a., Kaufmann V., Klopschitz M., Bischof H., Leberl F. (2010). Towards fully
automatic photogrammetric reconstruction using digital images taken from UAVs, IS-
PRS TC VII Symposium – 100 Years ISPRS, Vienna, Austria, IAPRS, Vol. XXXVIII,
Part 7A pp.65-70.
64. Issod, C.S. (2015). Getting Started with Hobby Quadcopters and Drones, USA, Mid-
dletown, DE. P.97.
65. Iwahashi J & Pike R. J (2007) “Automated classifications of topography from DEMs by
an unsupervised nested-means algorithm and a three-zone geometric signature,” Geo-
morphology, vol. 86, no. 3/4, pp. 409–440.
66. Jankowski, P. (1995). Integrating geographical information systems and multiple crite-
ria Decision-Making Methods. Geographical Information Systems 9(3) :251-273.
DOI:10.1080/02693799508902036
67. Jenness J. (2005) Topographic Position Index. Extension for ArcView 3.x.
http://jennessent.com.
68. Jenness J. (2010) “Topographic Position Index (tpi_jen.avx) extension for ArcView
3.x”, v. 1.3a. Jenness Enterprises, http://www.jennessent.com/arcview/tpi.htm.
69. Jenness J. (2004). "Calculating landscape surface area from digital elevation models."
Wildlife Society Bulletin 32.3, pp. 829-839.
70. Jenson S.K., Domingue, J.O., 1988. Extracting topographic structure from digital eleva-
tion model data for geographic information system analysis. Photogrammetric Engi-
neering and Remote Sensing. 54, pp. 1593–1600.
71. Kaaniche K., Champion, B., Pegard, C. and Vasseur, P. (2005). A vision algorithm for
dynamic detection of moving vehicles with a UAV, In: IEEE International Conference
on Robotics and Automation, pp. 1878-1883.
72. Karara H.M. (1989). Non-Topographic Photogrammetry. Americal Society for Photo-
grammetry and Remote Sensing.

105
73. Karara H.M., and W. Faig. (1980). An Expose on Photographic Data Aquisition Sys-
tems in Close-Range Photogrammetry. 14th Congress of the International Society for
Photogrammetry, Commission V, Hamburg, Germany.
74. Koeln G.T., Cowardin L.M., Strong L.L. (1996). Geographical information systems. in
T.A. Bookhout, ed. Research and management techniques for wildlife and habitats.
Fifth ed., rev. The Wildlife Society, Bethesda MD., pp. 540-566.
75. Lafay M. (2015). Drones for Dummies, John Wiley and sons, Inc. Hobokon, New Jersy.
P.272.
76. Lam, N. and Quattrochi, D.A. (1992). On the issues of scale, resolution, and fractal
analysis in the mapping sciences, Prof. Geogr., pp.44-88.
77. Leibowitz D., College M. (2001). Camera Calibration and Reconstruction of Geometry
from Images. Robotics Research Group Department of Engineering Science University
of Oxford Trinity Term, p. 209.
78. Levin S.A. (1992). The problem of pattern and scale in ecology. Ecology. 73, pp. 1943–
1967.
79. Li Z., (1993). Mathematical models of the accuracy of digital terrain model surfaces
linearly constructed from gridded data, Photogrammetric Record, (82), pp. 661-674.
80. Lowe, D. (2004). Distinctive image features from scale-invariant key points. Interna-
tional Journal of Computer Vision 60, pp. 91–110.
81. Lucieer A., Watson C. and Turner D. (2012) “Development of a UAV-LiDAR System
with Application to Forest Inventory Luke Wallace”. School of Geography and Envi-
ronmental Studies, University of Tasmania. http://www.mdpi.com/2072-4292/4/6/1519.
82. Luhmann, T, Robson, S, Kyle, S, and Harley, I. (2006). Close Range Photogrammetry:
Principles, Methods, and Applications. Caithness: Whittles Publishing.
83. Lundblad ER., Wright DJ., Miller J., Larkin EM., Rinehart R., Naar DF., Donahue BT.,
Anderson SM., Battista T A. (2006). benthic terrain classification scheme for American
Samoa. Mar Geod., 29, pp. 89–111.
84. MacMillan R.A, Shary P.A. (2009). Landforms and landform elements in geomor-
phometry. In: Hengl, T., Reuter, H.I. (Eds.), Geomorphometry—Concepts, Software,
Applications. Developments in Soil Science, vol. 33. Elsevier, Amsterdam, pp. 227–
254.
85. Mark D.M. (1975). Geomorphometric parameters: A review and evaluation, Ge-
ografiska Annaler. Ser. A, Phys. Geography, vol. 57, no. 3/4, pp. 165– 177.
86. Mark, D.M. and Aronson, P.B. (1984). Scale-dependent fractal dimensions of topo-
graphic surfaces: an empirical investigation, with applications in geomorphology and
computer mapping, Math. Geol., 11,671.
87. Marshall A. R. (1989). Network design and optimization in close ranges photogramme-
try.UNISURV S-36, School of Surveying, University of new South Wales, Australia., p.
249.
88. Masahiko N. (2007). UAV borne mapping system for river environment, In: 28th Asian
Association of Remote Sensing Conference, Kuala Lumpur, Malaysia
89. Matthews N. A. (2008). Aerial and Close-Range Photogrammetric Technology: Provid-
ing Resource Documentation, Interpretation, and Preservation. Technical Note 428.
U.S. Department of the Interior, Bureau of Land Management, National Operations
Center, Denver, Colorado. 42 pp.
90. McCormick M.I. (1994). Comparison of field methods for measuring surface topogra-
phy and their associations with a tropical reef fish assemblage. Marine Ecology Pro-
gress Series 112, pp. 87-96.

106
91. McGlone J.C. (1989). Analytic data-reduction schemes in non-topographic photogram-
metry, Non-Topographic Photogrammetry, 2nd ed., ed. H.M. Karara, ASPRS, Falls
Church, VA, Chap. 4, 37-55.
92. McHarg I. L. (1995). Design with Nature. New York: J. Wiley. (Reprinted from De-
sign)
93. Meyer R. (1987). “100 Years of Architectural Photogrammetry”, Kompendium Photo-
grametrie, Vol. XIX, Leipzig: Akademische Verlagsgesellschaft, pp. 183-200.
94. Micheletti N, Chandler JH, Lane SN. (2014). Investigating the geomorphological poten-
tial of freely available and accessible structure-from-motion photogrammetry using a
smartphone. Earth Surface Processes and Landforms, DOI: 10.1002/esp.3648.
95. Mikhail E., Bethel J. and McGlone J. (2001). Introduction to modern photogrammetry.
John Wiley & Sons, New York, p.479.
96. Moser K., Ahn, C., Noe, G. (2007) Characterization of micro topography and its influ-
ence on vegetation patterns in created wetlands. Wetlands 27, pp. 1081-1097.
97. Nagai M., Shibasaki R., Manandhar D. and Zhao H. (2004): Development of digital
surface and feature extraction by integrating laser scanner and CCD sensor with IMU,
In: International Archives of the Photogrammetry, Remote Sensing and Spatial Infor-
mation Sciences, XX ISPRS Congress, Istanbul, Turkey, XXXV-B5, pp.655-659.
98. Nellis M.D. and Briggs, J.M. (1989). the effect of spatial scale on konza landscape clas-
sification using textural analysis, landscape ecol.,2,93.
99. Newsome, Sam R. Jr. (2016). "Discrepancy Analysis Between Close-Range Photo-
grammetry and Terrestrial LiDAR" Electronic Theses & Dissertations. Paper 1423.
100. Nordberg K., Farnebäck, G., Forssén, P.-E., Granlund, G., Moe, A., Wiklund, J. and
Doherty, P. (2002). Vision for a UAV helicopter, In: Workshop on aerial robotics, Lau-
sanne, Switzerland.
101. Partsh J. (1911). Schlesien, eine Landeskunde fur das deutshe Volk.II, S.586, Breslau.
102. Patias, P., Saatsoglou-Paliadeli , C., Georgoula, O., Pateraki, M., Stamnas, A. and Kyr-
iakou, N.(2007). Photogrammetric documentation and digital representation of the
macedonian palace in Vergina-Aegeae, In: CIPA, XXI International CIPA Symposium,
Athens, Greece.
103. Pettersson, P.O. and Doherty, P. (2004). Probabilistic roadmap based path planning for
autonomous unmanned aerial vehicles, In: 14th Int’l Conf. on Automated Planning.
104. Peucker T.K., Douglas D.H., 1974. Detection of surface specific points by local parallel
processing of discrete terrain elevation data. Computer Graphics and Image Processing,
4, pp. 375-387.
105. Pike R. J. (2002). A bibliography of terrain modeling (geomorphometry), the quantita-
tive representation of topography - supplement 4.0., Open-File Rep. No. 02-465. U.S.
Geological Survey, Denver, p. 116.
106. Pogorelov A.V & Doumit J.A. (2009). Relief of Kuban river basin: Morphometric anal-
ysis. M.: Geoc, 208 p. (In Russian).
107. Puri A. (2004). A Survey of Unmanned Aerial Vehicles (UAV) for Traffic Surveil-
lance, Internal Report, Department of Computer Science and Engineering, University of
South Florida, Tampa, FL, USA, p. 29.
108. Rasemann S., Schmidt J., Schrott L., Dikau R. (2004). “Geomorphometry in mountain
terrain”. In: Bishop, M.P., Shroder, J.F. eds. GIS & Mountain Geomorphology. Spring-
er, Berlin, pp.101-145.
109. Redweik P. (2012). Photogrammetry - Springer. Retrieved from
http://link.springer.com/chapter/10.1007/978-3- 642-28000-9_4/fulltext.html.
110. Reg Austin (2010). Unmanned aircraft systems UAVs design development and deploy-
ment, John Wiley & Sons Ltd, p.332

107
111. Riga Declaration on remotely piloted aircraft (drones) " Farming the future of aviation"
Riga - 6 March 2015.
112. Riley S.J., De Gloria S. D., Elliot R. A. (1999). Terrain ruggedness index that quantifies
topographic heterogeneity. Intermountain journal of sciences, Vol.5, No. 1-4.
113. Risk, M.J. (1972). Fish diversity on a coral reef in the Virgin Islands. Atoll Research
Bulletin 193, pp. 1-6.
114. Roelofs R. (1951). Distortion, principal point, point of symmetry and calibrated princi-
pal point, Photogrammetria.vol.7, pp. 49-66.
115. Rüther H., Smit J., Kamamba D. (2012). A Comparison of Close-Range Photogramme-
try to Terrestrial Laser Scanning for Heritage Documentation// South African Journal of
Geomatics, Vol. 1, No. 2, pp. 149-169.
116. Shary P.A. (1995). Land surface in gravity points classification by a complete system of
curvatures. Mathematical Geology, 27(3), pp.373-390.
117. Shary P.A., Sharaya L.S., Mitusov A.V. (2002). Fundamental quantitative methods of
land surface analysis. Geoderma, 107(1-2), pp. 1-32.
118. SMITH, M. J. and PARK, D. W. G. (2000). Absolute and exterior orientation using linear
features. International Archives of Photogrammetry and Remote Sensing, 33(B3): 850-
857.
119. Snavely N., Seitz, S.M. & Szeliski, R. (2006). Photo Tourism: Exploring image collec-
tions in 3D. ACM Transactions on Graphics, 25, p.835.
120. Snavely, N., Seitz, S.M. & Szeliski, R. (2007). Modeling the World from Internet Photo
Collections. International Journal of Computer Vision, 80, pp. 189-210.
121. Snavely, N., Seitz, S.N., Szeliski, R. (2008). Modeling the world from internet photo
collections. International Journal of Computer Vision 80, pp.189-210.
122. Spetsakis M.E., Aloimonos Y. (1991). A multi-frame approach to visual motion percep-
tion. International Journal of Computer Vision 6.pp. 245-255.
123. Stambaugh M.C., Guyette R.P. (2008). Predicting spatio-temporal variability in fire
return intervals using a topographic roughness index. For Ecol Manag, 254(3), pp. 463–
473.
124. Stöcker, C., Bennett, R., Nex, F., Gerke, M., Schmidt, C., Zein, T. (2017). Guide to reg-
ulatory practice on UAVs for land tenure recording. H2020 its4land 687828 D4.1 inno-
vation for Land Tenure.p.48.
125. Stone R. O., Dugundji J. (1965). A study of micro relief-Its mapping, classification and
quantification by means of a Fourier analysis, Eng. Geol., vol. 1, no. 2, pp. 89-187.
126. Sugiura R., Noguchi, N. and Ishii, K. (2005). Remote-sensing Technology for Vegeta-
tion Monitoring using an Unmanned Helicopter, In: Biosystems Engineering, 90, 4,
369–379.
127. Szeliski R., Kang S.B. (1994). Recovering 3-D shape and motion from image streams
using nonlinear least squares. Journal of Visual Communication and Image Representa-
tion 5. pp.10-28.
128. Tagil S. & Jenness J (2008) GIS- based automated landform classification and topo-
graphic, Landcover, and geologic attributes of landforms around the Yazoren Polje,
Turkey. Journal of Applied Sciences 8(6): 910-921.
129. Tsai Y. R. (1986). An efficient and accurate camera calibration technique for 3D ma-
chine vision. In unknown focal length and principal point. In Proc. IEEE Conference on
Computer Vision.
130. Valentine P.C., Scully L.A., Fuller S.J. (2004). Terrain Ruggedness Analysis and Dis-
tribution of Boulder Ridges and Bedrock Outcrops in the Stellwagen Bank National
Marine Sanctuary Region —Posters presented at the Fifth International Symposium of
the Geological and Biological Habitat Mapping Group, Galway, Ireland.

108
131. Veitinger J., Purves R. S., Sovilla B. (2016). Potential slab avalanche release area iden-
tification from estimated winter terrain: a multi-scale, fuzzy logic approach/Natural
Hazards and Earth System Sciences, 16, pp. 2211–2225.
132. Verhoeven, G, Sevara, C, Karel, W, Ressl, C, Doneus, M and Briese, C. (2013) ‘Un-
distorting the past: new techniques for orthorectification of archaeological aerial frame
imagery’, in Corsi, C et al (eds) Good Practice in Archaeological Diagnostics, Natural
Science in Archaeology. Switzerland: Springer International Publishing.
133. Von Blyenburg P (1999) UAVs-Current Situation and Considerations for the Way For-
ward. RTO-AVT Course on Development and Operation of UAVs for Military and Civ-
il Applications.
134. Wedding L.M., Friedlander A.M., McGranaghan M., Yost R.S., Monaco M. E. (2008).
Using bathymetric LIDAR to define nearshore benthic habitat complexity: implications
for management of reef fish assemblages in Hawaii. Remote Sens Environ, 112(11), pp.
4159–4165.
135. Westoby M., Brasington J., Glasser N.F., Hambrey M.J., Reyonds M.J. (2012). Struc-
ture from Motion photogrammetry: a low-cost, effective tool for geoscience applica-
tions. Geomorphology 179, pp. 300-314.
136. Wolf P. R. and Ghilani C.D. (1997). Adjustment Computations: Statistics and Least
Squares in Surveying and GIS. John Wiley & Sons.
137. Wood J. (1996). The geomorphological characterization of digital elevation models,
PhD Thesis. University of Leicester.
138. Wood J. (2009). Geomorphometry in LandSerf. In: Hengl, T., Reuter, H.I. (Eds.), Geo-
morphometry — Concepts, Software, Applications. Developments in Soil Science, vol.
33. Elsevier, Amsterdam, pp. 333–349.
139. Woodby D., Carlile D., Hulbert L. (2009). Predictive modeling of coral distribution in
the Central Aleutian Islands, USA. Mar Ecol Prog Ser. 397, pp. 227–240.
140. Woodcock C. E., Strahler. A H. (1987). “The Factor of Scale in Remote Sensing,” Re-
mote Sensing of Environment, vol.21, pp. 311-332.
141. Wright D.J., Heyman W.D. (2008). Introduction to the special issue: marine and coastal
GIS for geomorphology, habitat mapping, and marine reserves. Mar Geod., 31, pp. 223–
230.
142. Wu J. (2004). Effects of changing scale on landscape pattern analysis: scaling relations
Landscape Ecol. 19, pp. 125–138.
143. Wzorek M., Land´en D. and Doherty P. (2006). GSM Technology as a Communication
Media for an Autonomous Unmanned Aerial Vehicle, In: 21th UAV Systems Confer-
ence, Bristol.
144. Xie Feifei et al. (2014). 1:500 Scale Aerial Triangulation Test with Unmanned Airship
in Hubei Province IOP Conf. Ser.: Earth Environ. Sci. 17 012177.
145. Zevenbergen L., Thorne C. (1987). Quantitative Analysis of Land Surface Topography,
Earth Surface Processes and Landforms, Vol. 12, pp. 47-56.
146. Zhou G., Li C. and Cheng P. (2005). Unmanned aerial vehicle (UAV) realtime video
registration for forest fire monitoring, In: IEEE International, Geoscience and Remote
Sensing Symposium, pp. 1803-1806.

109
 Scientific edition

FROM DRONES TO GEOSPATIAL DATA

Печатается в авторской редакции

Подписано в печать 26.12.2018. Формат 60×84 1/16.

Печать цифровая. Уч-изд. л.6,7. Тираж 60 экз.
Заказ № 3609.

Издательско-полиграфический центр КубГУ.

350040, г. Краснодар, ул. Ставропольская, 149.